{
  "$schema": "https://receiptaudit.com/benchmarks/schema.json",
  "suite": "receipt-audit",
  "title": "RA-Bench: receipt extraction, fake detection, evidence and duplicates",
  "preliminary": true,
  "disclosure": "We build Receipt Audit and we run this benchmark. Our engine is one of the targets, so read it as a vendor benchmark: the public split's cases, labels, every target's outputs and the code are open, every run is kept, and losses are listed first. To reproduce a number, run the command below with your own keys; to dispute one, email hello@godard.me.",
  "manifest": {
    "runId": "20261012T230424Z-fc67c6",
    "suite": "receipt-audit",
    "suiteVersion": "0.2.0",
    "datasetHash": "d36eb9d9f3c39095eb98d466bc8d846364ba8c6f867bac21dbdb7e52762a15d8",
    "split": "public",
    "startedAt": "2026-10-12T23:04:24.517Z",
    "finishedAt": "2026-10-12T23:18:58.299Z",
    "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5",
    "reps": 3,
    "cases": 150,
    "targets": [
      {
        "id": "self",
        "label": "Receipt Audit (Claude Haiku 5.5)",
        "kind": "self",
        "version": "engine.0.2.1+anthropic/claude-haiku-5.5+prompt.9a2a5877e979+git.3c7c53a7378b",
        "config": {
          "engine": "ReceiptChecks (apps/receipt-check/lib/server/ReceiptChecks.ts), in process",
          "engineVersion": "0.2.1",
          "production": true,
          "model": "anthropic/claude-haiku-5.5",
          "prompt": "apps/receipt-check/lib/receipts/model.ts (INSTRUCTIONS, USER_PROMPT and the ModelAnalysis output schema)",
          "promptSha256": "9a2a5877e9795c37df6d2c2cb89f56cae64b76ae10bbd672cc1ab6bb41426a71",
          "instructionsSha256": "da94bf411e9d3c47462a4d77b59a373a2b478a3e6afffb86889dc5e93ad452fb",
          "hashVersion": 2,
          "endpoint": "https://ai-gateway.vercel.sh/v1/chat/completions",
          "reasoningEffort": "low",
          "structuredOutput": true,
          "modelTimeout": "90 seconds",
          "modelRetries": "2, on errors effect/ai marks retryable",
          "filePreparation": "upright, at most 3072 px and 3.75 MB (lib/receipts/model.ts)",
          "duplicateNetwork": "the production Convex functions on an in-memory backend, empty at the start of each repetition",
          "submitter": "one account per case (the case id)",
          "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5",
          "pricePerCheckUsd": 0.01633333333333333,
          "cost": "cost_per_check_usd is what a caller pays Receipt Audit per check: the list price of the smallest credit pack (credits-3k: $49 for 3,000 checks, $0.0163 per check). model_cost_per_check_usd (and each result's costUsd) is the engine's own model call: tokens × the AI Gateway list price on the price date."
        },
        "priceDate": "2026-10-12",
        "unsupportedTasks": []
      },
      {
        "id": "self-gemini",
        "label": "Receipt Audit on Gemini 3.8 Flash",
        "kind": "self",
        "version": "engine.0.2.1+google/gemini-3.8-flash+prompt.9a2a5877e979+git.3c7c53a7378b",
        "config": {
          "engine": "ReceiptChecks (apps/receipt-check/lib/server/ReceiptChecks.ts), in process",
          "engineVersion": "0.2.1",
          "production": false,
          "model": "google/gemini-3.8-flash",
          "prompt": "apps/receipt-check/lib/receipts/model.ts (INSTRUCTIONS, USER_PROMPT and the ModelAnalysis output schema)",
          "promptSha256": "9a2a5877e9795c37df6d2c2cb89f56cae64b76ae10bbd672cc1ab6bb41426a71",
          "instructionsSha256": "da94bf411e9d3c47462a4d77b59a373a2b478a3e6afffb86889dc5e93ad452fb",
          "hashVersion": 2,
          "endpoint": "https://ai-gateway.vercel.sh/v1/chat/completions",
          "reasoningEffort": "low",
          "structuredOutput": true,
          "modelTimeout": "90 seconds",
          "modelRetries": "2, on errors effect/ai marks retryable",
          "filePreparation": "upright, at most 3072 px and 3.75 MB (lib/receipts/model.ts)",
          "duplicateNetwork": "the production Convex functions on an in-memory backend, empty at the start of each repetition",
          "submitter": "one account per case (the case id)",
          "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5",
          "pricePerCheckUsd": 0.01633333333333333,
          "cost": "cost_per_check_usd is what a caller pays Receipt Audit per check: the list price of the smallest credit pack (credits-3k: $49 for 3,000 checks, $0.0163 per check). model_cost_per_check_usd (and each result's costUsd) is the engine's own model call: tokens × the AI Gateway list price on the price date."
        },
        "priceDate": "2026-10-12",
        "unsupportedTasks": []
      },
      {
        "id": "diy:claude-haiku-5.5",
        "label": "Claude Haiku 5.5, plain prompt",
        "kind": "diy",
        "version": "claude-haiku-5-5@claude-code-cli+prompt.512dfb3893be+git.3c7c53a7378b",
        "config": {
          "transport": "claude-code-cli",
          "endpoint": "http://127.0.0.1:8789/v1/messages (claude-local: one headless `claude -p` per call, no tools, no MCP, no settings, empty cwd)",
          "billing": "subscription",
          "cli": "claude 2.1.296",
          "thinking": "Claude Code default (adaptive; can't be fully disabled via the CLI)",
          "ignored": [
            "max_tokens"
          ],
          "costUsd": "API-equivalent: the reported tokens × the AI Gateway list price, i.e. what the check costs when called through the API",
          "usage": "as the CLI reports it, including Claude Code's own cached system prompt (about 3,500 tokens, mostly cache reads); inputTokens counts uncached, cache-read and cache-write tokens",
          "model": "claude-haiku-5-5",
          "max_tokens": 4096,
          "message": "one user turn: the file (image block, or document block for a PDF), then the prompt",
          "rateLimits": "HTTP 429 (a subscription limit on claude-local) is retried with backoff like a transport error",
          "prompt": "apps/receipt-check/bench/prompt.ts (DIY_PLAIN_PROMPT)",
          "promptSha256": "512dfb3893be5aa7f89ff7be2259d8f9c94c948fe8b2d6e3dca9bc6f5889bc37",
          "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5",
          "cost": "cost_per_check_usd and model_cost_per_check_usd are both the API-equivalent token cost: tokens × the AI Gateway list price on the price date"
        },
        "priceDate": "2026-10-12",
        "unsupportedTasks": [
          "checks",
          "duplicates",
          "proof"
        ]
      },
      {
        "id": "diy:gemini-3.8-flash",
        "label": "Gemini 3.8 Flash, plain prompt",
        "kind": "diy",
        "version": "google/gemini-3.8-flash+prompt.512dfb3893be+git.3c7c53a7378b",
        "config": {
          "endpoint": "https://ai-gateway.vercel.sh/v1/chat/completions (Vercel AI Gateway, OpenAI-compatible)",
          "model": "google/gemini-3.8-flash",
          "max_tokens": "not set (provider default)",
          "reasoning": "not set (provider default)",
          "temperature": "not set (provider default)",
          "message": "one user turn: the file (image_url part, or a file part for a PDF), then the prompt",
          "prompt": "apps/receipt-check/bench/prompt.ts (DIY_PLAIN_PROMPT)",
          "promptSha256": "512dfb3893be5aa7f89ff7be2259d8f9c94c948fe8b2d6e3dca9bc6f5889bc37",
          "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5",
          "cost": "cost_per_check_usd and model_cost_per_check_usd are both the API-equivalent token cost: tokens × the AI Gateway list price on the price date"
        },
        "priceDate": "2026-10-12",
        "unsupportedTasks": [
          "checks",
          "duplicates",
          "proof"
        ]
      },
      {
        "id": "decision:gpt-6-luna",
        "label": "OpenAI decision model (gpt-6-luna), asked if it's fake",
        "kind": "diy",
        "version": "openai/gpt-6-luna-decisions+prompt.977e84771afc+git.3c7c53a7378b",
        "config": {
          "endpoint": "https://ai-gateway.vercel.sh/v1/decisions (Vercel AI Gateway → OpenAI Decisions API, public beta)",
          "model": "openai/gpt-6-luna-decisions",
          "questions": "apps/receipt-check/bench/prompt.ts (DECISION_QUESTIONS): a predicate (is it fake) and two choices (how it was made, which part looks altered)",
          "questionsSha256": "e17d0fb5b77b6147ffdb12fa1e22ed2dc99e5edec56d69731f297ad2c86aca37",
          "input": "one user message: DECISION_INPUT_TEXT and the image as a base64 data URL; a PDF's first page rendered at 150 dpi (pdftoppm)",
          "risk": "the predicate's probability",
          "reasons": "the two choices, in plain words",
          "cost": "input tokens × $0.1 per 1M (no output charge), price of 2026-10-12",
          "gitSha": "3c7c53a7378b5e1704315156cbbef1ae521b89a5"
        },
        "priceDate": "2026-10-12",
        "unsupportedTasks": [
          "extraction",
          "checks",
          "duplicates",
          "proof"
        ]
      }
    ]
  },
  "metrics": [
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "all",
      "target": "self",
      "value": 1,
      "lo": 0.9213484012671117,
      "hi": 1,
      "n": 45,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "all",
      "target": "self",
      "value": 1,
      "lo": 0.9123783988027135,
      "hi": 1,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "all",
      "target": "self",
      "value": 1,
      "lo": 0.9229264369144198,
      "hi": 1,
      "n": 46,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "all",
      "target": "self",
      "value": 0.9971264367816093,
      "lo": 0.962596275098718,
      "hi": 0.9997863132665838,
      "n": 116,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "all",
      "target": "self",
      "value": 0.9916666666666667,
      "lo": 0.9543037087549505,
      "hi": 0.998527444266658,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "all",
      "target": "self",
      "value": 0.9722222222222221,
      "lo": 0.9251355540193794,
      "hi": 0.9900130110296227,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "all",
      "target": "self",
      "value": 0.9166666666666666,
      "lo": 0.741511978067893,
      "hi": 0.9768411847029567,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "all",
      "target": "self",
      "value": 0.9444444444444445,
      "lo": 0.8880247010855891,
      "hi": 0.9732915954311192,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "all",
      "target": "self",
      "value": 0.9308510638297872,
      "lo": 0.8740956885259877,
      "hi": 0.9709467166650619,
      "n": 98,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "all",
      "target": "self",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 1,
      "lo": 0.8864866068260312,
      "hi": 0.9999999999999999,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.996376811594203,
      "lo": 0.9532321227978544,
      "hi": 0.9997305559548253,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9895833333333334,
      "lo": 0.9433338237728424,
      "hi": 0.9981588297831838,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9756944444444443,
      "lo": 0.9220728923934436,
      "hi": 0.9927107497141845,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9122807017543859,
      "lo": 0.7077914971685695,
      "hi": 0.9780958093143608,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9444444444444445,
      "lo": 0.8792692439659896,
      "hi": 0.9754191223827436,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9420468557336621,
      "lo": 0.8856323375303875,
      "hi": 0.9815969060308454,
      "n": 81,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.9761904761904762,
      "lo": 0.749095974501273,
      "hi": 0.9982270705105024,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.8,
      "lo": 0.6756756756756757,
      "hi": 0.9565217391304348,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "base=cord",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "base=cord",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "base=cord",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "base=cord",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.9545454545454546,
      "lo": 0.8331521739130439,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "base=cord",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "capture=photo",
      "target": "self",
      "value": 1,
      "lo": 0.9059421741627975,
      "hi": 1,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "capture=photo",
      "target": "self",
      "value": 1,
      "lo": 0.895730002457548,
      "hi": 1,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "capture=photo",
      "target": "self",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9969418960244649,
      "lo": 0.9602764446288532,
      "hi": 0.9997725875754877,
      "n": 109,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9910714285714286,
      "lo": 0.9511515316911152,
      "hi": 0.9984221452967079,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9702380952380951,
      "lo": 0.9199927306857199,
      "hi": 0.9892960024298922,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.8823529411764705,
      "lo": 0.6566364940511411,
      "hi": 0.9671202262362073,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9404761904761906,
      "lo": 0.880374853071378,
      "hi": 0.9713639602267907,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9376470588235294,
      "lo": 0.8788353556268199,
      "hi": 0.9775993005771118,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "capture=photo",
      "target": "self",
      "value": 1,
      "lo": 0.9668386529330562,
      "hi": 1,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8063923194655637,
      "hi": 1,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.9166666666666666,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.9166666666666666,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.9227642276422764,
      "lo": 0.8398602033405955,
      "hi": 0.9871835443037975,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=genuine",
      "target": "self",
      "value": 1,
      "lo": 0.8241207763533425,
      "hi": 1,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=genuine",
      "target": "self",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=genuine",
      "target": "self",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.995169082125604,
      "lo": 0.9384697313335618,
      "hi": 0.9996407134550588,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=genuine",
      "target": "self",
      "value": 1,
      "lo": 0.9472627418109483,
      "hi": 1,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.9806763285024154,
      "lo": 0.9148053639606295,
      "hi": 0.9958481897610065,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.9758454106280192,
      "lo": 0.9074744724784117,
      "hi": 0.9940267842208964,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.9365853658536586,
      "lo": 0.8383027694610778,
      "hi": 0.9916146211363119,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=genuine",
      "target": "self",
      "value": 1,
      "lo": 0.9472627418109483,
      "hi": 1,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.9090909090909093,
      "lo": 0.6226415635484046,
      "hi": 0.9837678271141153,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.8181818181818182,
      "lo": 0.523019438039156,
      "hi": 0.9486323102539149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.7272727272727273,
      "lo": 0.43435469882387084,
      "hi": 0.902539407099751,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.9090909090909091,
      "lo": 0.7878539493293591,
      "hi": 0.986490990990991,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.9230769230769231,
      "lo": 0.7272727272727273,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.9955555555555556,
      "lo": 0.9431509048812243,
      "hi": 0.9996694645691506,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.9688888888888888,
      "lo": 0.9014262781928679,
      "hi": 0.9906593628565896,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.951111111111111,
      "lo": 0.8767030738757225,
      "hi": 0.9815594149064569,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.9368421052631579,
      "lo": 0.864702818627451,
      "hi": 0.9840169002702001,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.9615384615384616,
      "lo": 0.8110722109696524,
      "hi": 0.9931780150064564,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.935897435897436,
      "lo": 0.7755182780748768,
      "hi": 0.9840513797914483,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 1,
      "lo": 0.916201209134296,
      "hi": 0.9999999999999999,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 1,
      "lo": 0.9179901967742088,
      "hi": 1,
      "n": 43,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9970501474926254,
      "lo": 0.9616360914658402,
      "hi": 0.9997806390948507,
      "n": 113,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9914529914529915,
      "lo": 0.9531704967258107,
      "hi": 0.9984896464262035,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9715099715099714,
      "lo": 0.9232863704396193,
      "hi": 0.9897556799468059,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9047619047619048,
      "lo": 0.7108586092608176,
      "hi": 0.9734812325619255,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9430199430199431,
      "lo": 0.8852734104134612,
      "hi": 0.9725999360221227,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9290393013100436,
      "lo": 0.8695421294383552,
      "hi": 0.9706904078266013,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 1,
      "lo": 0.9682107543372667,
      "hi": 1,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "editedField=total",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "editedField=total",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "editedField=total",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.9722222222222222,
      "lo": 0.7174912732783316,
      "hi": 0.9979310537694965,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.9,
      "lo": 0.7619047619047619,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "editedField=total",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.8668077490609515,
      "hi": 0.9999999999999999,
      "n": 25,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.967741935483871,
      "lo": 0.838058948352751,
      "hi": 0.9942827784745831,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.9354838709677421,
      "lo": 0.7928136326370189,
      "hi": 0.9821252164780855,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.8712871287128713,
      "lo": 0.7578867623604466,
      "hi": 0.9615530303030303,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.9955555555555556,
      "lo": 0.9431509048812243,
      "hi": 0.9996694645691506,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.9688888888888888,
      "lo": 0.9014262781928679,
      "hi": 0.9906593628565896,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.951111111111111,
      "lo": 0.8767030738757225,
      "hi": 0.9815594149064569,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.9368421052631579,
      "lo": 0.8641720480691101,
      "hi": 0.9840637760062221,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=id",
      "target": "self",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.9215686274509803,
      "lo": 0.8031967213114756,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=free",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.8717948717948717,
      "lo": 0.6061982305249418,
      "hi": 0.9677821336219359,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.9166666666666666,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.935,
      "lo": 0.8452339034205232,
      "hi": 0.9947916666666666,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=none",
      "target": "self",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=none",
      "target": "self",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=none",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.9958333333333333,
      "lo": 0.9465402263251748,
      "hi": 0.9996901285520938,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.9875,
      "lo": 0.9325372992233133,
      "hi": 0.9977900244795476,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.9583333333333333,
      "lo": 0.8896831488976221,
      "hi": 0.984983565694811,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.9416666666666667,
      "lo": 0.8673348107018559,
      "hi": 0.9755258415417618,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.9332191780821918,
      "lo": 0.8469472605694565,
      "hi": 0.9839471578171963,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=none",
      "target": "self",
      "value": 1,
      "lo": 0.9541818704644728,
      "hi": 0.9999999999999999,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.9696969696969698,
      "lo": 0.6985052183558994,
      "hi": 0.9977426005529196,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.9833333333333334,
      "lo": 0.8121528279044053,
      "hi": 0.9987595193112028,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.9166666666666666,
      "lo": 0.8174603174603174,
      "hi": 0.9818181818181818,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.9944444444444446,
      "lo": 0.9297988116076135,
      "hi": 0.9995868012740978,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.9833333333333333,
      "lo": 0.9114487027240993,
      "hi": 0.9970518402052136,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.961111111111111,
      "lo": 0.8784211269356963,
      "hi": 0.9883092760888206,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.9388888888888888,
      "lo": 0.8480606054973967,
      "hi": 0.9768996576223241,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.9332191780821918,
      "lo": 0.8568993781129444,
      "hi": 0.9839937840519183,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=cord",
      "target": "self",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.9230769230769231,
      "lo": 0.7272727272727273,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.8666666666666666,
      "lo": 0.5589289935914743,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.8666666666666666,
      "lo": 0.5589289935914743,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.875,
      "lo": 0.75,
      "hi": 0.9824561403508771,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.9885057471264368,
      "lo": 0.8635858763751989,
      "hi": 0.9991447802629856,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.9655172413793104,
      "lo": 0.8282447812067973,
      "hi": 0.9938867857072374,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.9540229885057472,
      "lo": 0.81174756337716,
      "hi": 0.9900844586747997,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.9128919860627178,
      "lo": 0.7610665977298674,
      "hi": 0.9983952463463369,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.9135802469135802,
      "lo": 0.7503195676549773,
      "hi": 0.9738139262622437,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.9135802469135802,
      "lo": 0.7503195676549773,
      "hi": 0.9738139262622437,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.9457364341085271,
      "lo": 0.8900057273768613,
      "hi": 0.9824561403508771,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "all",
      "target": "self",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "all",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.05202306241026692,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "all",
      "target": "self",
      "value": 0.6944444444444445,
      "lo": 0.6070595250933093,
      "hi": 0.7697663546327507,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.05202306241026692,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.732638888888889,
      "lo": 0.6364143535749367,
      "hi": 0.8109615881857284,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0.4,
      "lo": 0.16818032970623614,
      "hi": 0.6873262302663417,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.7857142857142857,
      "lo": 0.524107694133997,
      "hi": 0.9242861328730683,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 3.469446951953614e-18,
      "hi": 0.05273725818905173,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.7261904761904763,
      "lo": 0.6371422043539997,
      "hi": 0.8002371862585734,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.6833333333333333,
      "lo": 0.46474800533957883,
      "hi": 0.8428394367077209,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.7053140096618358,
      "lo": 0.5892319744696185,
      "hi": 0.7997406489793216,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.2775327998628892,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.45454545454545453,
      "lo": 0.21271271622459764,
      "hi": 0.719908462590678,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.688888888888889,
      "lo": 0.5770831354521158,
      "hi": 0.7822878573877134,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.22809537235419838,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.7307692307692307,
      "lo": 0.5391695962082381,
      "hi": 0.8629555167798162,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.05202306241026692,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.6951566951566952,
      "lo": 0.6066419972163251,
      "hi": 0.7712636248469387,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.2153108027376358,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.7419354838709677,
      "lo": 0.5675386419056621,
      "hi": 0.8629829409360625,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.688888888888889,
      "lo": 0.5770831354521158,
      "hi": 0.7822878573877134,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.7666666666666667,
      "lo": 0.4578711470372018,
      "hi": 0.927444693035924,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.6,
      "lo": 0.31267376973365824,
      "hi": 0.8318196702937638,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.7833333333333334,
      "lo": 0.5661276081063968,
      "hi": 0.9092348023303392,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.05202306241026692,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.6708333333333334,
      "lo": 0.5621168189838291,
      "hi": 0.7638953200915325,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.7,
      "lo": 0.4810271816464765,
      "hi": 0.8545227551323956,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.6611111111111112,
      "lo": 0.5349046316195614,
      "hi": 0.7679288826902098,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.2153108027376358,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.7,
      "lo": 0.4810271816464765,
      "hi": 0.8545227551323956,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.6666666666666667,
      "lo": 0.3677978121758822,
      "hi": 0.8730245878698215,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.9885057471264368,
      "lo": 0.8635858763751989,
      "hi": 0.9991447802629856,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.25925925925925924,
      "lo": 0.1317037456190719,
      "hi": 0.44678571314553067,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "all",
      "target": "self",
      "value": 0.9375,
      "lo": 0.880625,
      "hi": 0.98437890625,
      "n": 120,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "all",
      "target": "self",
      "value": 0.85,
      "lo": 0.725,
      "hi": 0.95,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "all",
      "target": "self",
      "value": 0.85,
      "lo": 0.725,
      "hi": 0.975,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "all",
      "target": "self",
      "value": 0.9,
      "lo": 0.775,
      "hi": 0.975,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "all",
      "target": "self",
      "value": 0.6,
      "lo": 0.44595893660346186,
      "hi": 0.7365167431570808,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "all",
      "target": "self",
      "value": 0.5416666666666667,
      "lo": 0.390480657117095,
      "hi": 0.6855508761164645,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.380952380952381,
      "lo": 0.17979560866977065,
      "hi": 0.6333736300772856,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.3095238095238095,
      "lo": 0.13213685461323155,
      "hi": 0.5689339273820583,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.9263521288837745,
      "lo": 0.855379746835443,
      "hi": 0.9825517836593786,
      "n": 112,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.8484848484848485,
      "lo": 0.7272727272727273,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.8484848484848485,
      "lo": 0.7272727272727273,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.8787878787878788,
      "lo": 0.7575757575757576,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.5858585858585859,
      "lo": 0.4175982352384029,
      "hi": 0.7362139874058224,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.5252525252525253,
      "lo": 0.3613496687833472,
      "hi": 0.6838892202296604,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.9222222222222223,
      "lo": 0.8083333333333333,
      "hi": 1,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.4222222222222222,
      "lo": 0.21449779087720927,
      "hi": 0.6616618337283673,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.35555555555555557,
      "lo": 0.16686720968758836,
      "hi": 0.603143521722768,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 26,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.9344594594594594,
      "lo": 0.8694130067567567,
      "hi": 0.9844594594594595,
      "n": 117,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.8648648648648649,
      "lo": 0.7297297297297297,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.8648648648648649,
      "lo": 0.7567567567567568,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.8918918918918919,
      "lo": 0.7837837837837838,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.5945945945945946,
      "lo": 0.43485964730229365,
      "hi": 0.7365348180798572,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.5315315315315317,
      "lo": 0.3754979927677672,
      "hi": 0.6816334956929501,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.5555555555555555,
      "lo": 0.29701927644834225,
      "hi": 0.7871480561455196,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.4444444444444445,
      "lo": 0.21285194385448025,
      "hi": 0.7029807235516576,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.9824561403508771,
      "lo": 0.9342105263157895,
      "hi": 1,
      "n": 31,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.8333333333333334,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.638888888888889,
      "lo": 0.36629477503065894,
      "hi": 0.8441235564539966,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.5833333333333334,
      "lo": 0.31951131254954973,
      "hi": 0.8067396863412435,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.9222222222222223,
      "lo": 0.8055555555555556,
      "hi": 1,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.4222222222222222,
      "lo": 0.21449779087720927,
      "hi": 0.6616618337283673,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.35555555555555557,
      "lo": 0.16686720968758836,
      "hi": 0.603143521722768,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.6666666666666667,
      "lo": 0.3677978121758822,
      "hi": 0.8730245878698215,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.6,
      "lo": 0.31267376973365824,
      "hi": 0.8318196702937638,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.7666666666666666,
      "lo": 0.45787114703720166,
      "hi": 0.9274446930359239,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.7333333333333333,
      "lo": 0.42677693205514144,
      "hi": 0.9103744280088436,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.4333333333333334,
      "lo": 0.190125658185213,
      "hi": 0.7135453817965056,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.3333333333333333,
      "lo": 0.12697541213017843,
      "hi": 0.632202187824118,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.8666666666666668,
      "lo": 0.5589289935914745,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.5333333333333334,
      "lo": 0.26109856576627416,
      "hi": 0.7870659142428669,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.9,
      "lo": 0.6989663547715128,
      "hi": 0.9721335187862319,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.8833333333333334,
      "lo": 0.6786711249982054,
      "hi": 0.9644662538279667,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.29999999999999993,
      "lo": 0.14547724486760424,
      "hi": 0.5189728183535234,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.19999999999999998,
      "lo": 0.080657662579798,
      "hi": 0.4160174322518936,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.93375,
      "lo": 0.8484296874999999,
      "hi": 0.996875,
      "n": 100,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.8,
      "lo": 0.65,
      "hi": 0.95,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.8,
      "lo": 0.65,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.7166666666666667,
      "lo": 0.4975386865533784,
      "hi": 0.8659737449570666,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.6666666666666667,
      "lo": 0.4486927772355472,
      "hi": 0.8309321700801799,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.94125,
      "lo": 0.821875,
      "hi": 1,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.5333333333333333,
      "lo": 0.26109856576627405,
      "hi": 0.7870659142428666,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.5,
      "lo": 0.236593090512564,
      "hi": 0.7634069094874361,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.94125,
      "lo": 0.818125,
      "hi": 1,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.4333333333333334,
      "lo": 0.190125658185213,
      "hi": 0.7135453817965056,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.3333333333333333,
      "lo": 0.12697541213017843,
      "hi": 0.632202187824118,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.9038461538461539,
      "lo": 0.8163281250000001,
      "hi": 0.975,
      "n": 106,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.7692307692307693,
      "lo": 0.6153846153846154,
      "hi": 0.9230769230769231,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.7692307692307693,
      "lo": 0.6153846153846154,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.8461538461538461,
      "lo": 0.6538461538461539,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.6153846153846153,
      "lo": 0.42534850085115516,
      "hi": 0.775714055642872,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.5897435897435896,
      "lo": 0.40133251536474096,
      "hi": 0.7550494730195023,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 94,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.5714285714285714,
      "lo": 0.3259064457965074,
      "hi": 0.7861920109552588,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.45238095238095244,
      "lo": 0.2314529890368834,
      "hi": 0.6938147064619392,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "all",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.703186733179637,
      "hi": 0.9469034451594526,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "all",
      "target": "self",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.703186733179637,
      "hi": 0.9469034451594526,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.07134759913335872,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.875,
      "lo": 0.6899611872949993,
      "hi": 0.9565566591987655,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.20833333333333334,
      "lo": 0.09244825352851857,
      "hi": 0.40470453253188665,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06313883485297894,
      "n": 57,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.703186733179637,
      "hi": 0.9469034451594526,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.2588329669680317,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.2588329669680317,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.75,
      "lo": 0.5050168346449122,
      "hi": 0.8981793250878696,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "currency=USD",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.1936076805344365,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.703186733179637,
      "hi": 0.9469034451594526,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "lang=en",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.1843181350088518,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.75,
      "lo": 0.5050168346449122,
      "hi": 0.8981793250878696,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.8666666666666667,
      "lo": 0.703186733179637,
      "hi": 0.9469034451594526,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.75,
      "lo": 0.5050168346449122,
      "hi": 0.8981793250878696,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.3,
      "lo": 0.10779126740630099,
      "hi": 0.6032218525388546,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.154639018924847,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 2.7755575615628914e-17,
      "hi": 0.24249400665524085,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "decision_consistency",
      "stratum": "all",
      "target": "self",
      "value": 0.9583333333333334,
      "lo": 0.9061591476629647,
      "hi": 0.9820732831198906,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "total_consistency",
      "stratum": "all",
      "target": "self",
      "value": 0.975,
      "lo": 0.9290699666721339,
      "hi": 0.9914618252300978,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "calibration_error",
      "stratum": "all",
      "target": "self",
      "value": 0.08470277777777771,
      "lo": 0.06140173611111109,
      "hi": 0.12305069444444432,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "all",
      "target": "self",
      "value": 400,
      "lo": 258.33333333333354,
      "hi": 541.874999999999,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "all",
      "target": "self",
      "value": 458.33333333333326,
      "lo": 308.33333333333337,
      "hi": 608.3333333333334,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self",
      "value": 40,
      "lo": 25.83333333333334,
      "hi": 54.16666666666668,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self",
      "value": 45.83333333333333,
      "lo": 30.83333333333334,
      "hi": 60.02083333333323,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_5pct",
      "target": "self",
      "value": 2000.0000000000005,
      "lo": 1291.6666666666667,
      "hi": 2708.333333333334,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_5pct",
      "target": "self",
      "value": 2291.6666666666665,
      "lo": 1541.666666666667,
      "hi": 3083.333333333334,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_32usd",
      "target": "self",
      "value": 128,
      "lo": 80.00000000000004,
      "hi": 173.33333333333331,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_32usd",
      "target": "self",
      "value": 146.66666666666663,
      "lo": 101.33333333333333,
      "hi": 194.66666666666669,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_182usd",
      "target": "self",
      "value": 728,
      "lo": 484.9541666666672,
      "hi": 971.0458333333314,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_182usd",
      "target": "self",
      "value": 834.1666666666665,
      "lo": 576.3333333333334,
      "hi": 1122.3333333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_8usd",
      "target": "self",
      "value": 400,
      "lo": 250.0000000000001,
      "hi": 533.3333333333334,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_8usd",
      "target": "self",
      "value": 458.33333333333326,
      "lo": 308.33333333333337,
      "hi": 600.0000000000001,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_73usd",
      "target": "self",
      "value": 400,
      "lo": 258.3333333333334,
      "hi": 550,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_73usd",
      "target": "self",
      "value": 458.33333333333326,
      "lo": 308.33333333333337,
      "hi": 600,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "all",
      "target": "self",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self",
      "value": 100,
      "lo": 100,
      "hi": 100,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_5pct",
      "target": "self",
      "value": 5000,
      "lo": 5000,
      "hi": 5000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_32usd",
      "target": "self",
      "value": 320,
      "lo": 320,
      "hi": 320,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_182usd",
      "target": "self",
      "value": 1820,
      "lo": 1820,
      "hi": 1820,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_8usd",
      "target": "self",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_73usd",
      "target": "self",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "evidence",
      "metric": "edited_field_named",
      "stratum": "all",
      "target": "self",
      "value": 0.3,
      "lo": 0.1454772448676043,
      "hi": 0.5189728183535235,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "all",
      "target": "self",
      "value": 0.4666666666666667,
      "lo": 0.32188304157682207,
      "hi": 0.6172917318363305,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.5,
      "lo": 0.2992980081982123,
      "hi": 0.7007019918017877,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.5333333333333333,
      "lo": 0.26109856576627405,
      "hi": 0.7870659142428666,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.3333333333333333,
      "lo": 0.12697541213017843,
      "hi": 0.632202187824118,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "genuine_with_proof_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "all",
      "target": "self",
      "value": 6299,
      "lo": 6299,
      "hi": 6299,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "all",
      "target": "self",
      "value": 9951.249999999996,
      "lo": 9951.249999999996,
      "hi": 9951.249999999996,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "all",
      "target": "self",
      "value": 12417.549999999996,
      "lo": 12417.549999999996,
      "hi": 12417.549999999996,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "all",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "all",
      "target": "self",
      "value": 0.0010725777777777779,
      "lo": 0.0010725777777777779,
      "hi": 0.0010725777777777779,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "all",
      "target": "self",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 6641.5,
      "lo": 6641.5,
      "hi": 6641.5,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 10046.949999999999,
      "lo": 10046.949999999999,
      "hi": 10046.949999999999,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 12304.150000000007,
      "lo": 12304.150000000007,
      "hi": 12304.150000000007,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0.0010786386243386249,
      "lo": 0.0010786386243386249,
      "hi": 0.0010786386243386249,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=consistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 7382.5,
      "lo": 7382.5,
      "hi": 7382.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 9772.2,
      "lo": 9772.2,
      "hi": 9772.2,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 9940.07,
      "lo": 9940.07,
      "hi": 9940.07,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0.0011352619047619048,
      "lo": 0.0011352619047619048,
      "hi": 0.0011352619047619048,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 4298.5,
      "lo": 4298.5,
      "hi": 4298.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 6189.849999999995,
      "lo": 6189.849999999995,
      "hi": 6189.849999999995,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 12716.310000000007,
      "lo": 12716.310000000007,
      "hi": 12716.310000000007,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0.0009084533333333333,
      "lo": 0.0009084533333333333,
      "hi": 0.0009084533333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=unknown",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=cord",
      "target": "self",
      "value": 6821.5,
      "lo": 6821.5,
      "hi": 6821.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=cord",
      "target": "self",
      "value": 9733.199999999999,
      "lo": 9733.199999999999,
      "hi": 9733.199999999999,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=cord",
      "target": "self",
      "value": 9940.07,
      "lo": 9940.07,
      "hi": 9940.07,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=cord",
      "target": "self",
      "value": 0.001084688095238095,
      "lo": 0.001084688095238095,
      "hi": 0.001084688095238095,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 6844,
      "lo": 6844,
      "hi": 6844,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 9334.55,
      "lo": 9334.55,
      "hi": 9334.55,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 9686.109999999999,
      "lo": 9686.109999999999,
      "hi": 9686.109999999999,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 0.001092588888888889,
      "lo": 0.001092588888888889,
      "hi": 0.001092588888888889,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=photo",
      "target": "self",
      "value": 6474.5,
      "lo": 6474.5,
      "hi": 6474.5,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=photo",
      "target": "self",
      "value": 9958.749999999998,
      "lo": 9958.749999999998,
      "hi": 9958.749999999998,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=photo",
      "target": "self",
      "value": 12587.650000000003,
      "lo": 12587.650000000003,
      "hi": 12587.650000000003,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0.0010727500000000004,
      "lo": 0.0010727500000000004,
      "hi": 0.0010727500000000004,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=photo",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=scan",
      "target": "self",
      "value": 6120,
      "lo": 6120,
      "hi": 6120,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=scan",
      "target": "self",
      "value": 7760.5,
      "lo": 7760.5,
      "hi": 7760.5,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=scan",
      "target": "self",
      "value": 9519.91,
      "lo": 9519.91,
      "hi": 9519.91,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "self",
      "value": 0.0010875375,
      "lo": 0.0010875375,
      "hi": 0.0010875375,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=scan",
      "target": "self",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=scan",
      "target": "self",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=scan",
      "target": "self",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 5004,
      "lo": 5004,
      "hi": 5004,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 9589,
      "lo": 9589,
      "hi": 9589,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 9964.199999999999,
      "lo": 9964.199999999999,
      "hi": 9964.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 0.0010535999999999998,
      "lo": 0.0010535999999999998,
      "hi": 0.0010535999999999998,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 7068.5,
      "lo": 7068.5,
      "hi": 7068.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 9913.75,
      "lo": 9913.75,
      "hi": 9913.75,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 10326.259999999998,
      "lo": 10326.259999999998,
      "hi": 10326.259999999998,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0.0011631166666666664,
      "lo": 0.0011631166666666664,
      "hi": 0.0011631166666666664,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_full",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 6451,
      "lo": 6451,
      "hi": 6451,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 9186.8,
      "lo": 9186.8,
      "hi": 9186.8,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 9624.07,
      "lo": 9624.07,
      "hi": 9624.07,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0.0010738833333333332,
      "lo": 0.0010738833333333332,
      "hi": 0.0010738833333333332,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_inpaint",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 6750,
      "lo": 6750,
      "hi": 6750,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 9326.35,
      "lo": 9326.35,
      "hi": 9326.35,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 10363.16,
      "lo": 10363.16,
      "hi": 10363.16,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0.0010897744444444443,
      "lo": 0.0010897744444444443,
      "hi": 0.0010897744444444443,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=duplicate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=genuine",
      "target": "self",
      "value": 5295,
      "lo": 5295,
      "hi": 5295,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=genuine",
      "target": "self",
      "value": 10566.799999999996,
      "lo": 10566.799999999996,
      "hi": 10566.799999999996,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=genuine",
      "target": "self",
      "value": 13627.459999999997,
      "lo": 13627.459999999997,
      "hi": 13627.459999999997,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0.0010496893719806764,
      "lo": 0.0010496893719806764,
      "hi": 0.0010496893719806764,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=genuine",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 4160,
      "lo": 4160,
      "hi": 4160,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 10263,
      "lo": 10263,
      "hi": 10263,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 10605.36,
      "lo": 10605.36,
      "hi": 10605.36,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0.0009783060606060607,
      "lo": 0.0009783060606060607,
      "hi": 0.0009783060606060607,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=hard_negative",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 7180,
      "lo": 7180,
      "hi": 7180,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 9760.55,
      "lo": 9760.55,
      "hi": 9760.55,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 9972.83,
      "lo": 9972.83,
      "hi": 9972.83,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0.001100233333333333,
      "lo": 0.001100233333333333,
      "hi": 0.001100233333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=manual_edit",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 4399.5,
      "lo": 4399.5,
      "hi": 4399.5,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 7200.05,
      "lo": 7200.05,
      "hi": 7200.05,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 7232.01,
      "lo": 7232.01,
      "hi": 7232.01,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 0.001028588888888889,
      "lo": 0.001028588888888889,
      "hi": 0.001028588888888889,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=EUR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 6847,
      "lo": 6847,
      "hi": 6847,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 10379.399999999998,
      "lo": 10379.399999999998,
      "hi": 10379.399999999998,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 12994.519999999973,
      "lo": 12994.519999999973,
      "hi": 12994.519999999973,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0.0010926886446886446,
      "lo": 0.0010926886446886446,
      "hi": 0.0010926886446886446,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=IDR",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=USD",
      "target": "self",
      "value": 5203,
      "lo": 5203,
      "hi": 5203,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=USD",
      "target": "self",
      "value": 9043.05,
      "lo": 9043.05,
      "hi": 9043.05,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=USD",
      "target": "self",
      "value": 9719.85,
      "lo": 9719.85,
      "hi": 9719.85,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0.0010046333333333332,
      "lo": 0.0010046333333333332,
      "hi": 0.0010046333333333332,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=USD",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 6348,
      "lo": 6348,
      "hi": 6348,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 9985,
      "lo": 9985,
      "hi": 9985,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 12454.00000000001,
      "lo": 12454.00000000001,
      "hi": 12454.00000000001,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0.0010734514739229022,
      "lo": 0.0010734514739229022,
      "hi": 0.0010734514739229022,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "docType=receipt",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 4554,
      "lo": 4554,
      "hi": 4554,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 9020.65,
      "lo": 9020.65,
      "hi": 9020.65,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 9517.73,
      "lo": 9517.73,
      "hi": 9517.73,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 0.0009870444444444442,
      "lo": 0.0009870444444444442,
      "hi": 0.0009870444444444442,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=items+total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=total",
      "target": "self",
      "value": 7499,
      "lo": 7499,
      "hi": 7499,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=total",
      "target": "self",
      "value": 9775.25,
      "lo": 9775.25,
      "hi": 9775.25,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=total",
      "target": "self",
      "value": 9956.449999999999,
      "lo": 9956.449999999999,
      "hi": 9956.449999999999,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0.0011534388888888888,
      "lo": 0.0011534388888888888,
      "hi": 0.0011534388888888888,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=total",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=en",
      "target": "self",
      "value": 5207,
      "lo": 5207,
      "hi": 5207,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=en",
      "target": "self",
      "value": 8914,
      "lo": 8914,
      "hi": 8914,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=en",
      "target": "self",
      "value": 9677.099999999999,
      "lo": 9677.099999999999,
      "hi": 9677.099999999999,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=en",
      "target": "self",
      "value": 0.001011751111111111,
      "lo": 0.001011751111111111,
      "hi": 0.001011751111111111,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=en",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=id",
      "target": "self",
      "value": 6847,
      "lo": 6847,
      "hi": 6847,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=id",
      "target": "self",
      "value": 10379.399999999998,
      "lo": 10379.399999999998,
      "hi": 10379.399999999998,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=id",
      "target": "self",
      "value": 12994.519999999973,
      "lo": 12994.519999999973,
      "hi": 12994.519999999973,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=id",
      "target": "self",
      "value": 0.0010926886446886446,
      "lo": 0.0010926886446886446,
      "hi": 0.0010926886446886446,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=id",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 7192.5,
      "lo": 7192.5,
      "hi": 7192.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 9224.8,
      "lo": 9224.8,
      "hi": 9224.8,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 9808.390000000001,
      "lo": 9808.390000000001,
      "hi": 9808.390000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0.0012039,
      "lo": 0.0012039,
      "hi": 0.0012039,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=free",
      "target": "self",
      "value": 5691.5,
      "lo": 5691.5,
      "hi": 5691.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=free",
      "target": "self",
      "value": 9984.25,
      "lo": 9984.25,
      "hi": 9984.25,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=free",
      "target": "self",
      "value": 10532.060000000001,
      "lo": 10532.060000000001,
      "hi": 10532.060000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0.001122333333333333,
      "lo": 0.001122333333333333,
      "hi": 0.001122333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=free",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 6912.5,
      "lo": 6912.5,
      "hi": 6912.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 9789.3,
      "lo": 9789.3,
      "hi": 9789.3,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 10330.389999999998,
      "lo": 10330.389999999998,
      "hi": 10330.389999999998,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0.0011420533333333332,
      "lo": 0.0011420533333333332,
      "hi": 0.0011420533333333332,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=intact",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=none",
      "target": "self",
      "value": 5806.5,
      "lo": 5806.5,
      "hi": 5806.5,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=none",
      "target": "self",
      "value": 10190.900000000001,
      "lo": 10190.900000000001,
      "hi": 10190.900000000001,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=none",
      "target": "self",
      "value": 12694.099999999999,
      "lo": 12694.099999999999,
      "hi": 12694.099999999999,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0.0010534833333333334,
      "lo": 0.0010534833333333334,
      "hi": 0.0010534833333333334,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 7035,
      "lo": 7035,
      "hi": 7035,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 9410.249999999998,
      "lo": 9410.249999999998,
      "hi": 9410.249999999998,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 9965.35,
      "lo": 9965.35,
      "hi": 9965.35,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0.0011081216666666666,
      "lo": 0.0011081216666666666,
      "hi": 0.0011081216666666666,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=stripped",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=cord",
      "target": "self",
      "value": 6795,
      "lo": 6795,
      "hi": 6795,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=cord",
      "target": "self",
      "value": 10600.049999999996,
      "lo": 10600.049999999996,
      "hi": 10600.049999999996,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=cord",
      "target": "self",
      "value": 13426.06999999999,
      "lo": 13426.06999999999,
      "hi": 13426.06999999999,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=cord",
      "target": "self",
      "value": 0.0010910820175438602,
      "lo": 0.0010910820175438602,
      "hi": 0.0010910820175438602,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=cord",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 6785,
      "lo": 6785,
      "hi": 6785,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 8328.6,
      "lo": 8328.6,
      "hi": 8328.6,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 8469.72,
      "lo": 8469.72,
      "hi": 8469.72,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.0010364266666666668,
      "lo": 0.0010364266666666668,
      "hi": 0.0010364266666666668,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 5148,
      "lo": 5148,
      "hi": 5148,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 9412.099999999999,
      "lo": 9412.099999999999,
      "hi": 9412.099999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 9701.619999999999,
      "lo": 9701.619999999999,
      "hi": 9701.619999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 0.00111134,
      "lo": 0.00111134,
      "hi": 0.00111134,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 7180,
      "lo": 7180,
      "hi": 7180,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 9760.55,
      "lo": 9760.55,
      "hi": 9760.55,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 9972.83,
      "lo": 9972.83,
      "hi": 9972.83,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0.001100233333333333,
      "lo": 0.001100233333333333,
      "hi": 0.001100233333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 4738,
      "lo": 4738,
      "hi": 4738,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 8527.599999999999,
      "lo": 8527.599999999999,
      "hi": 8527.599999999999,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 9487.99,
      "lo": 9487.99,
      "hi": 9487.99,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0.0009694392156862743,
      "lo": 0.0009694392156862743,
      "hi": 0.0009694392156862743,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=expressexpense",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 7160,
      "lo": 7160,
      "hi": 7160,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 10018,
      "lo": 10018,
      "hi": 10018,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 10532.060000000001,
      "lo": 10532.060000000001,
      "hi": 10532.060000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0.0011486833333333335,
      "lo": 0.0011486833333333335,
      "hi": 0.0011486833333333335,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 7035,
      "lo": 7035,
      "hi": 7035,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 8649.15,
      "lo": 8649.15,
      "hi": 8649.15,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 8933.57,
      "lo": 8933.57,
      "hi": 8933.57,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0.00117755,
      "lo": 0.00117755,
      "hi": 0.00117755,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 7531,
      "lo": 7531,
      "hi": 7531,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 10895,
      "lo": 10895,
      "hi": 10895,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 13531.560000000003,
      "lo": 13531.560000000003,
      "hi": 13531.560000000003,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0.0011586957264957268,
      "lo": 0.0011586957264957268,
      "hi": 0.0011586957264957268,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=exclusive",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 5064,
      "lo": 5064,
      "hi": 5064,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 9373.399999999996,
      "lo": 9373.399999999996,
      "hi": 9373.399999999996,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 11962.079999999998,
      "lo": 11962.079999999998,
      "hi": 11962.079999999998,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0.001028312121212121,
      "lo": 0.001028312121212121,
      "hi": 0.001028312121212121,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=none",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=crop",
      "target": "self",
      "value": 5857,
      "lo": 5857,
      "hi": 5857,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=crop",
      "target": "self",
      "value": 10373.8,
      "lo": 10373.8,
      "hi": 10373.8,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=crop",
      "target": "self",
      "value": 10405.16,
      "lo": 10405.16,
      "hi": 10405.16,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "self",
      "value": 0.0010309666666666664,
      "lo": 0.0010309666666666664,
      "hi": 0.0010309666666666664,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=crop",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=crop",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=crop",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 5743,
      "lo": 5743,
      "hi": 5743,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 7755.499999999999,
      "lo": 7755.499999999999,
      "hi": 7755.499999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 8195.1,
      "lo": 8195.1,
      "hi": 8195.1,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 0.0010555866666666664,
      "lo": 0.0010555866666666664,
      "hi": 0.0010555866666666664,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=jpeg",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 6722,
      "lo": 6722,
      "hi": 6722,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 8864.599999999999,
      "lo": 8864.599999999999,
      "hi": 8864.599999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 9364.119999999999,
      "lo": 9364.119999999999,
      "hi": 9364.119999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 0.0011280666666666664,
      "lo": 0.0011280666666666664,
      "hi": 0.0011280666666666664,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rephoto",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=resize",
      "target": "self",
      "value": 6902,
      "lo": 6902,
      "hi": 6902,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=resize",
      "target": "self",
      "value": 8643.199999999999,
      "lo": 8643.199999999999,
      "hi": 8643.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=resize",
      "target": "self",
      "value": 8841.44,
      "lo": 8841.44,
      "hi": 8841.44,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "self",
      "value": 0.0010530799999999999,
      "lo": 0.0010530799999999999,
      "hi": 0.0010530799999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=resize",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=resize",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=resize",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 8062,
      "lo": 8062,
      "hi": 8062,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 9024.9,
      "lo": 9024.9,
      "hi": 9024.9,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 9205.78,
      "lo": 9205.78,
      "hi": 9205.78,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 0.0012173466666666668,
      "lo": 0.0012173466666666668,
      "hi": 0.0012173466666666668,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rotate",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 5004,
      "lo": 5004,
      "hi": 5004,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 9589,
      "lo": 9589,
      "hi": 9589,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 9964.199999999999,
      "lo": 9964.199999999999,
      "hi": 9964.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 0.0010535999999999998,
      "lo": 0.0010535999999999998,
      "hi": 0.0010535999999999998,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=screenshot",
      "target": "self",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9777777777777777,
      "lo": 0.8843331823158513,
      "hi": 0.9960664011171667,
      "n": 45,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "all",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9123783988027135,
      "hi": 1,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "all",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9229264369144198,
      "hi": 1,
      "n": 46,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9971264367816093,
      "lo": 0.962596275098718,
      "hi": 0.9997863132665838,
      "n": 116,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9972222222222222,
      "lo": 0.9638041693022279,
      "hi": 0.9997934374258393,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9833333333333333,
      "lo": 0.941264037027841,
      "hi": 0.9954174354340789,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9722222222222223,
      "lo": 0.8177838751262339,
      "hi": 0.9963497093473961,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9666666666666667,
      "lo": 0.917419662440894,
      "hi": 0.9869624489016494,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.940552016985138,
      "lo": 0.8892161799261497,
      "hi": 0.9825123297569336,
      "n": 98,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "all",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.9705882352941176,
      "lo": 0.8508442670731406,
      "hi": 0.9947890937237309,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8864866068260312,
      "hi": 0.9999999999999999,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9599186107143782,
      "hi": 1,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.9965277777777777,
      "lo": 0.955105373991343,
      "hi": 0.9997417852888818,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.9649122807017545,
      "lo": 0.7780617091912259,
      "hi": 0.9953856789703768,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.9791666666666665,
      "lo": 0.9271928648221578,
      "hi": 0.9942680301475696,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.964329643296433,
      "lo": 0.9282994678277697,
      "hi": 0.9938070396801708,
      "n": 81,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.9285714285714286,
      "lo": 0.6853129557584889,
      "hi": 0.9872777847521093,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.9285714285714286,
      "lo": 0.6853129557584889,
      "hi": 0.9872777847521093,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.45,
      "hi": 0.9583333333333334,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.9285714285714286,
      "lo": 0.6853129557584889,
      "hi": 0.9872777847521093,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.9285714285714286,
      "lo": 0.6853129557584889,
      "hi": 0.9872777847521093,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.9545454545454546,
      "lo": 0.8333333333333334,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.972972972972973,
      "lo": 0.861759301362695,
      "hi": 0.9952130255480596,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.895730002457548,
      "hi": 1,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9969418960244649,
      "lo": 0.9602764446288532,
      "hi": 0.9997725875754877,
      "n": 109,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9970238095238095,
      "lo": 0.9613049808095904,
      "hi": 0.9997786801417214,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9821428571428571,
      "lo": 0.9372195125788385,
      "hi": 0.9950891884637512,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9607843137254901,
      "lo": 0.7568661674622448,
      "hi": 0.9948406492943036,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9642857142857143,
      "lo": 0.9117533615468005,
      "hi": 0.9860253876053231,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9589201877934272,
      "lo": 0.9223334685872726,
      "hi": 0.9886384996582365,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9668386529330562,
      "hi": 1,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8063923194655637,
      "hi": 1,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.9666666666666668,
      "lo": 0.7873326750830553,
      "hi": 0.9956171774009802,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.9666666666666668,
      "lo": 0.7873326750830553,
      "hi": 0.9956171774009802,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.8821138211382114,
      "lo": 0.7261838624338625,
      "hi": 0.9887702185454994,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.9444444444444444,
      "lo": 0.7424269921812741,
      "hi": 0.990124809021697,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.995169082125604,
      "lo": 0.9384697313335618,
      "hi": 0.9996407134550588,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.9951690821256037,
      "lo": 0.9384697313335614,
      "hi": 0.9996407134550588,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.9855072463768116,
      "lo": 0.9223687753885529,
      "hi": 0.9974370753554117,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.9661835748792269,
      "lo": 0.8933540361260528,
      "hi": 0.9898426265285991,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.9590643274853801,
      "lo": 0.9005701754385965,
      "hi": 1,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9472627418109483,
      "hi": 1,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.963963963963964,
      "lo": 0.919047619047619,
      "hi": 0.9954545454545455,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9615384615384616,
      "lo": 0.8110722109696524,
      "hi": 0.9931780150064564,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9871794871794871,
      "lo": 0.8498847742592102,
      "hi": 0.9990460198266822,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9871794871794872,
      "lo": 0.8498847742592105,
      "hi": 0.9990460198266823,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9615384615384616,
      "lo": 0.8110722109696524,
      "hi": 0.9931780150064564,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9102564102564104,
      "lo": 0.7420976017868207,
      "hi": 0.9727914879697207,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9761904761904762,
      "lo": 0.8767879842623334,
      "hi": 0.9957845958655673,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9179901967742088,
      "hi": 1,
      "n": 43,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9970501474926254,
      "lo": 0.9616360914658402,
      "hi": 0.9997806390948507,
      "n": 113,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9971509971509972,
      "lo": 0.9629057439597568,
      "hi": 0.999788139982426,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9829059829059829,
      "lo": 0.9398099145713432,
      "hi": 0.9952996173954185,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9682539682539684,
      "lo": 0.7958607110560711,
      "hi": 0.9958265569349456,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9658119658119658,
      "lo": 0.9153821514157074,
      "hi": 0.9866261581805496,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9389978213507625,
      "lo": 0.8853746790757382,
      "hi": 0.980600163519684,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9682107543372667,
      "hi": 1,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.9166666666666666,
      "lo": 0.646120088858883,
      "hi": 0.9851349055950829,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.9166666666666666,
      "lo": 0.646120088858883,
      "hi": 0.9851349055950829,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.7619047619047619,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.967741935483871,
      "lo": 0.838058948352751,
      "hi": 0.9942827784745831,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8668077490609515,
      "hi": 0.9999999999999999,
      "n": 25,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.9892473118279569,
      "lo": 0.8714103181530285,
      "hi": 0.9991999938157919,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.989247311827957,
      "lo": 0.8714103181530287,
      "hi": 0.9991999938157919,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.967741935483871,
      "lo": 0.838058948352751,
      "hi": 0.9942827784745831,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.924731182795699,
      "lo": 0.7785690051121716,
      "hi": 0.9772355514321893,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.8627450980392157,
      "lo": 0.7096774193548387,
      "hi": 0.9696969696969697,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.963963963963964,
      "lo": 0.9230676328502416,
      "hi": 0.9956714621585312,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.8627450980392157,
      "lo": 0.6274509803921569,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.9487179487179487,
      "lo": 0.6994900818041261,
      "hi": 0.9932448404421059,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.9166666666666666,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.8855721393034826,
      "lo": 0.7076666666666668,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9473684210526315,
      "lo": 0.753612688331012,
      "hi": 0.9906480059376995,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9958333333333333,
      "lo": 0.9465402263251748,
      "hi": 0.9996901285520938,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9958333333333333,
      "lo": 0.9465402263251748,
      "hi": 0.9996901285520938,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9875,
      "lo": 0.9325372992233133,
      "hi": 0.9977900244795476,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9708333333333334,
      "lo": 0.9072753191609728,
      "hi": 0.9912459421930726,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.9641025641025641,
      "lo": 0.9148936170212766,
      "hi": 1,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9541818704644728,
      "hi": 0.9999999999999999,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.9230769230769231,
      "lo": 0.8204545454545455,
      "hi": 1,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.9641025641025641,
      "lo": 0.9148905947775628,
      "hi": 1,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.9473684210526315,
      "lo": 0.753612688331012,
      "hi": 0.9906480059376995,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.9833333333333332,
      "lo": 0.812152827904405,
      "hi": 0.9987595193112027,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.9833333333333334,
      "lo": 0.8121528279044053,
      "hi": 0.9987595193112028,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.8833333333333334,
      "lo": 0.6786711249982054,
      "hi": 0.9644662538279667,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.8083333333333333,
      "lo": 0.562962962962963,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0.96875,
      "lo": 0.9042553191489362,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0.9534883720930233,
      "lo": 0.8555482456140351,
      "hi": 1,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.8709677419354839,
      "lo": 0.5690288068436802,
      "hi": 0.9718374424359149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.0047619047619047615,
      "lo": 0.00035415231289029,
      "hi": 0.060697261883945516,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.7027777777777778,
      "lo": 0.6157481744199719,
      "hi": 0.7772273858657762,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.0047619047619047615,
      "lo": 0.00035415231289029,
      "hi": 0.060697261883945516,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.7395833333333334,
      "lo": 0.6437894471878516,
      "hi": 0.8169410002970126,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.4333333333333333,
      "lo": 0.19012565818521288,
      "hi": 0.7135453817965055,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.7857142857142857,
      "lo": 0.524107694133997,
      "hi": 0.9242861328730683,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.8709677419354839,
      "lo": 0.5690288068436802,
      "hi": 0.9718374424359149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.004830917874396135,
      "lo": 0.00035928654494127993,
      "hi": 0.061530268666438445,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.7351190476190477,
      "lo": 0.6465884447168566,
      "hi": 0.8080559218409498,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.4810271816464765,
      "hi": 0.8545227551323956,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.7149758454106281,
      "lo": 0.599352389877729,
      "hi": 0.8079248276158674,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.2775327998628892,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0.45454545454545453,
      "lo": 0.21271271622459764,
      "hi": 0.719908462590678,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.6933333333333334,
      "lo": 0.5816954517744293,
      "hi": 0.7861313291322192,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.02564102564102564,
      "lo": 0.001909533775505512,
      "hi": 0.2657706912784776,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.7435897435897437,
      "lo": 0.5524602347728494,
      "hi": 0.8720051622700972,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.6428571428571429,
      "lo": 0.38764423103907397,
      "hi": 0.8365526824644587,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.precision",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8709677419354839,
      "lo": 0.5690288068436802,
      "hi": 0.9718374424359149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.0047619047619047615,
      "lo": 0.00035415231289029,
      "hi": 0.060697261883945516,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.7037037037037038,
      "lo": 0.615557169697166,
      "hi": 0.7788990635513502,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "totals_add_up.recall",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.023809523809523808,
      "lo": 0.00177292948949756,
      "hi": 0.25090402549872703,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.7526881720430109,
      "lo": 0.5788824347431106,
      "hi": 0.8707734406693574,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.6933333333333334,
      "lo": 0.5816954517744293,
      "hi": 0.7861313291322192,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.6,
      "lo": 0.31267376973365824,
      "hi": 0.8318196702937638,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.5839825677481064,
      "hi": 0.919342337420202,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.0047619047619047615,
      "lo": 0.00035415231289029,
      "hi": 0.060697261883945516,
      "n": 70,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.6791666666666667,
      "lo": 0.570702272909494,
      "hi": 0.7712128973402755,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.4810271816464765,
      "hi": 0.8545227551323956,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.06419393671876342,
      "n": 56,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.6666666666666666,
      "lo": 0.5405686645211968,
      "hi": 0.7727073847647731,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.023809523809523808,
      "lo": 0.00177292948949756,
      "hi": 0.25090402549872703,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.7166666666666667,
      "lo": 0.4975386865533784,
      "hi": 0.8659737449570666,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "checks",
      "metric": "false_alarm_rate",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "checks",
      "metric": "coverage",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0.25925925925925924,
      "lo": 0.1317037456190719,
      "hi": 0.44678571314553067,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.94375,
      "lo": 0.8887343750000001,
      "hi": 0.9871875,
      "n": 120,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.6,
      "lo": 0.45,
      "hi": 0.975,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.8,
      "hi": 0.975,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.8,
      "hi": 0.975,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.8333333333333334,
      "lo": 0.6900089677569106,
      "hi": 0.9182432981115651,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.016666666666666666,
      "lo": 0.003579607073928591,
      "hi": 0.07404458481041426,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.825,
      "lo": 0.6805000966821229,
      "hi": 0.9125458625396408,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.619047619047619,
      "lo": 0.36662636992271436,
      "hi": 0.8202043913302293,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.619047619047619,
      "lo": 0.36662636992271436,
      "hi": 0.8202043913302293,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.9370924434215574,
      "lo": 0.8741848868431147,
      "hi": 0.984656693517453,
      "n": 112,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.8787878787878788,
      "lo": 0.7575757575757576,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.8787878787878788,
      "lo": 0.7575757575757576,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.8787878787878788,
      "lo": 0.7575757575757576,
      "hi": 0.9696969696969697,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.8080808080808081,
      "lo": 0.6447979150790949,
      "hi": 0.9071165308795962,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.004219409282700422,
      "lo": 0.00031379493234937397,
      "hi": 0.05410491704516887,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.7979797979797979,
      "lo": 0.6335882962261585,
      "hi": 0.9002305941273294,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.8983333333333333,
      "lo": 0.7933333333333333,
      "hi": 1,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.9930555555555556,
      "lo": 0.9583333333333334,
      "hi": 1,
      "n": 26,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.07407407407407407,
      "lo": 0.01603905376269879,
      "hi": 0.2819321367511207,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.05555555555555555,
      "lo": 0.009875190978303097,
      "hi": 0.25757300781872594,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.9395270270270271,
      "lo": 0.8859712837837838,
      "hi": 0.9859797297297297,
      "n": 117,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.5945945945945946,
      "lo": 0.4594594594594595,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8918918918918919,
      "lo": 0.7837837837837838,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8918918918918919,
      "lo": 0.7837837837837838,
      "hi": 0.972972972972973,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8288288288288289,
      "lo": 0.678314232930216,
      "hi": 0.9174855753029755,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.016666666666666666,
      "lo": 0.003579607073928591,
      "hi": 0.07404458481041426,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8198198198198198,
      "lo": 0.6680884441078861,
      "hi": 0.9113880817079573,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.9868421052631579,
      "lo": 0.9473684210526315,
      "hi": 1,
      "n": 31,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.75,
      "lo": 0.5,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.75,
      "lo": 0.5,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.9722222222222222,
      "lo": 0.7174912732783316,
      "hi": 0.9979310537694965,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.07017543859649122,
      "lo": 0.015186418206402347,
      "hi": 0.26973958141777654,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.9722222222222222,
      "lo": 0.7174912732783316,
      "hi": 0.9979310537694965,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.05263157894736842,
      "lo": 0.009351994062300362,
      "hi": 0.24638731166898786,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.8983333333333333,
      "lo": 0.79,
      "hi": 1,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 0.934999999999991,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.06666666666666667,
      "lo": 0.014419867104976586,
      "hi": 0.2585552698741336,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.05,
      "lo": 0.008881448800795402,
      "hi": 0.23613119344674205,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.8666666666666668,
      "lo": 0.5589289935914745,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.8666666666666668,
      "lo": 0.5589289935914745,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.8833333333333332,
      "lo": 0.6786711249982051,
      "hi": 0.9644662538279665,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.8833333333333332,
      "lo": 0.6786711249982051,
      "hi": 0.9644662538279665,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.7833333333333333,
      "lo": 0.5661276081063967,
      "hi": 0.909234802330339,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.7666666666666666,
      "lo": 0.5485724655436072,
      "hi": 0.8988274501615557,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.016666666666666666,
      "lo": 0.003579607073928591,
      "hi": 0.07404458481041426,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.969375,
      "lo": 0.9118671875,
      "hi": 1,
      "n": 100,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.6,
      "lo": 0.4,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.95,
      "lo": 0.85,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.95,
      "lo": 0.85,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.9333333333333333,
      "lo": 0.7414447301258665,
      "hi": 0.9855801328950234,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.9166666666666667,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.94625,
      "lo": 0.83875,
      "hi": 1,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.6,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.7,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.89,
      "lo": 0.7375,
      "hi": 1,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.4,
      "lo": 0.1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.5,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.5,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.9139423076923077,
      "lo": 0.8341225961538462,
      "hi": 0.9798076923076923,
      "n": 106,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.4230769230769231,
      "lo": 0.2692307692307692,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.8461538461538461,
      "lo": 0.6923076923076923,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.8461538461538461,
      "lo": 0.6923076923076923,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.7435897435897437,
      "lo": 0.5524602347728494,
      "hi": 0.8720051622700972,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.730769230769231,
      "lo": 0.5391695962082385,
      "hi": 0.8629555167798163,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.9991071428571429,
      "lo": 0.9946428571428572,
      "hi": 1,
      "n": 94,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.9285714285714286,
      "lo": 0.7857142857142857,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.8555555555555555,
      "lo": 0.6900674297582453,
      "hi": 0.9403230462069324,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.8555555555555555,
      "lo": 0.6900674297582453,
      "hi": 0.9403230462069324,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 6.938893903907228e-18,
      "hi": 0.07134759913335872,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.861111111111111,
      "lo": 0.6735028863579292,
      "hi": 0.9490698547101405,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.20833333333333334,
      "lo": 0.09244825352851857,
      "hi": 0.40470453253188665,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06313883485297894,
      "n": 57,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0.8555555555555555,
      "lo": 0.6900674297582453,
      "hi": 0.9403230462069324,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.2588329669680317,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.2588329669680317,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.7291666666666666,
      "lo": 0.4842920979345887,
      "hi": 0.8853043818204612,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.1936076805344365,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.8555555555555555,
      "lo": 0.6900674297582453,
      "hi": 0.9403230462069324,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.1843181350088518,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.7291666666666666,
      "lo": 0.4842920979345887,
      "hi": 0.8853043818204612,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.8555555555555555,
      "lo": 0.6900674297582453,
      "hi": 0.9403230462069324,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.2,
      "lo": 0.0950510717728987,
      "hi": 0.3730569641314825,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.14865487460862087,
      "n": 22,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.062117855787202914,
      "n": 58,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.7291666666666666,
      "lo": 0.4842920979345887,
      "hi": 0.8853043818204612,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.1875,
      "lo": 0.06591599071428142,
      "hi": 0.4300888096197414,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08762160119728664,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.21428571428571427,
      "lo": 0.07571386712693184,
      "hi": 0.47589230586600295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "recall",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "recall_exact",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0.3,
      "lo": 0.10779126740630099,
      "hi": 0.6032218525388546,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.154639018924847,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 2.7755575615628914e-17,
      "hi": 0.24249400665524085,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "duplicates",
      "metric": "false_match_rate_genuine",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "decision_consistency",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.975,
      "lo": 0.9290699666721339,
      "hi": 0.9914618252300978,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "total_consistency",
      "stratum": "all",
      "target": "self-gemini",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "calibration_error",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.023488888888888854,
      "lo": 0.01889416666666681,
      "hi": 0.0698543749999998,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "all",
      "target": "self-gemini",
      "value": 447.16666666666663,
      "lo": 108.33333333333317,
      "hi": 1016.5354166666667,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "all",
      "target": "self-gemini",
      "value": 385.375,
      "lo": 91.66666666666679,
      "hi": 872.7916666666665,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self-gemini",
      "value": 299.71666666666664,
      "lo": 12.5,
      "hi": 801.7208333333332,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self-gemini",
      "value": 229.78750000000002,
      "lo": 9.166666666666679,
      "hi": 664.3833333333331,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_5pct",
      "target": "self-gemini",
      "value": 1102.4999999999998,
      "lo": 483.9583333333329,
      "hi": 1846.2812499999989,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_5pct",
      "target": "self-gemini",
      "value": 1076.8750000000002,
      "lo": 458.3333333333328,
      "hi": 1855.6249999999998,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_32usd",
      "target": "self-gemini",
      "value": 333.83333333333337,
      "lo": 37.33333333333331,
      "hi": 878.8999999999996,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_32usd",
      "target": "self-gemini",
      "value": 266.375,
      "lo": 31.999999999999993,
      "hi": 705.7916666666666,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_182usd",
      "target": "self-gemini",
      "value": 583.8333333333334,
      "lo": 197.16666666666637,
      "hi": 1116.458333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_182usd",
      "target": "self-gemini",
      "value": 528.8750000000001,
      "lo": 181.99999999999997,
      "hi": 1069,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_8usd",
      "target": "self-gemini",
      "value": 298.66666666666663,
      "lo": 107.99999999999996,
      "hi": 554.3333333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_8usd",
      "target": "self-gemini",
      "value": 274,
      "lo": 99.99999999999997,
      "hi": 546,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_73usd",
      "target": "self-gemini",
      "value": 1371.166666666667,
      "lo": 108.3333333333334,
      "hi": 3504.4583333333308,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_73usd",
      "target": "self-gemini",
      "value": 1078.375,
      "lo": 91.66666666666679,
      "hi": 2985.1249999999995,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "all",
      "target": "self-gemini",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_0_1pct",
      "target": "self-gemini",
      "value": 100,
      "lo": 100,
      "hi": 100,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_5pct",
      "target": "self-gemini",
      "value": 5000,
      "lo": 5000,
      "hi": 5000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_32usd",
      "target": "self-gemini",
      "value": 320,
      "lo": 320,
      "hi": 320,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_182usd",
      "target": "self-gemini",
      "value": 1820,
      "lo": 1820,
      "hi": 1820,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_8usd",
      "target": "self-gemini",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_73usd",
      "target": "self-gemini",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "evidence",
      "metric": "edited_field_named",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.6333333333333334,
      "lo": 0.4172262990322165,
      "hi": 0.8064736588203651,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.6,
      "lo": 0.44595893660346186,
      "hi": 0.7365167431570808,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.5,
      "lo": 0.2992980081982123,
      "hi": 0.7007019918017877,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "fakes_caught_with_proof",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.5,
      "lo": 0.236593090512564,
      "hi": 0.7634069094874361,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "proof",
      "metric": "genuine_with_proof_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.004166666666666667,
      "lo": 0.0003098714479059343,
      "hi": 0.053459773674825115,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "all",
      "target": "self-gemini",
      "value": 4518,
      "lo": 4518,
      "hi": 4518,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "all",
      "target": "self-gemini",
      "value": 12345.199999999975,
      "lo": 12345.199999999975,
      "hi": 12345.199999999975,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "all",
      "target": "self-gemini",
      "value": 25765.859999999946,
      "lo": 25765.859999999946,
      "hi": 25765.859999999946,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0.004379346666666661,
      "lo": 0.004379346666666661,
      "hi": 0.004379346666666661,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "all",
      "target": "self-gemini",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 4489.5,
      "lo": 4489.5,
      "hi": 4489.5,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 11100.499999999996,
      "lo": 11100.499999999996,
      "hi": 11100.499999999996,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 24145.780000000104,
      "lo": 24145.780000000104,
      "hi": 24145.780000000104,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0.00432952182539682,
      "lo": 0.00432952182539682,
      "hi": 0.00432952182539682,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=consistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 5030.5,
      "lo": 5030.5,
      "hi": 5030.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 19277.649999999987,
      "lo": 19277.649999999987,
      "hi": 19277.649999999987,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 37274.979999999894,
      "lo": 37274.979999999894,
      "hi": 37274.979999999894,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0.004816571428571429,
      "lo": 0.004816571428571429,
      "hi": 0.004816571428571429,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 4322.5,
      "lo": 4322.5,
      "hi": 4322.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 7383.549999999999,
      "lo": 7383.549999999999,
      "hi": 7383.549999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 9691.330000000002,
      "lo": 9691.330000000002,
      "hi": 9691.330000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0.0043950249999999995,
      "lo": 0.0043950249999999995,
      "hi": 0.0043950249999999995,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=unknown",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 4929,
      "lo": 4929,
      "hi": 4929,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 10726.499999999995,
      "lo": 10726.499999999995,
      "hi": 10726.499999999995,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 37101.54999999989,
      "lo": 37101.54999999989,
      "hi": 37101.54999999989,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0.004288928571428572,
      "lo": 0.004288928571428572,
      "hi": 0.004288928571428572,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 4579,
      "lo": 4579,
      "hi": 4579,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 16839.499999999996,
      "lo": 16839.499999999996,
      "hi": 16839.499999999996,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 19253.499999999993,
      "lo": 19253.499999999993,
      "hi": 19253.499999999993,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 0.005132416666666667,
      "lo": 0.005132416666666667,
      "hi": 0.005132416666666667,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 4490,
      "lo": 4490,
      "hi": 4490,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 11705.099999999979,
      "lo": 11705.099999999979,
      "hi": 11705.099999999979,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 27988.92000000006,
      "lo": 27988.92000000006,
      "hi": 27988.92000000006,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0.004366704044117642,
      "lo": 0.004366704044117642,
      "hi": 0.004366704044117642,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=photo",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 4772,
      "lo": 4772,
      "hi": 4772,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 15341.449999999993,
      "lo": 15341.449999999993,
      "hi": 15341.449999999993,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 21207.759999999995,
      "lo": 21207.759999999995,
      "hi": 21207.759999999995,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 0.0045615625000000005,
      "lo": 0.0045615625000000005,
      "hi": 0.0045615625000000005,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=scan",
      "target": "self-gemini",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 4786,
      "lo": 4786,
      "hi": 4786,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 9183.699999999997,
      "lo": 9183.699999999997,
      "hi": 9183.699999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 10846.339999999998,
      "lo": 10846.339999999998,
      "hi": 10846.339999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 0.00444875,
      "lo": 0.00444875,
      "hi": 0.00444875,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 4677,
      "lo": 4677,
      "hi": 4677,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 8354.249999999998,
      "lo": 8354.249999999998,
      "hi": 8354.249999999998,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 13917.16999999999,
      "lo": 13917.16999999999,
      "hi": 13917.16999999999,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0.004566425,
      "lo": 0.004566425,
      "hi": 0.004566425,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_full",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 4733.5,
      "lo": 4733.5,
      "hi": 4733.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 18259.49999999999,
      "lo": 18259.49999999999,
      "hi": 18259.49999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 40817.620000000024,
      "lo": 40817.620000000024,
      "hi": 40817.620000000024,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0.0048579750000000005,
      "lo": 0.0048579750000000005,
      "hi": 0.0048579750000000005,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_inpaint",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 4586.5,
      "lo": 4586.5,
      "hi": 4586.5,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 12143.099999999995,
      "lo": 12143.099999999995,
      "hi": 12143.099999999995,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 25573.139999999985,
      "lo": 25573.139999999985,
      "hi": 25573.139999999985,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0.0043253250000000005,
      "lo": 0.0043253250000000005,
      "hi": 0.0043253250000000005,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=duplicate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 4456,
      "lo": 4456,
      "hi": 4456,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 12221.699999999973,
      "lo": 12221.699999999973,
      "hi": 12221.699999999973,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 28042.39999999998,
      "lo": 28042.39999999998,
      "hi": 28042.39999999998,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0.004326528985507245,
      "lo": 0.004326528985507245,
      "hi": 0.004326528985507245,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=genuine",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 4311,
      "lo": 4311,
      "hi": 4311,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 7291.599999999995,
      "lo": 7291.599999999995,
      "hi": 7291.599999999995,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 11224.24,
      "lo": 11224.24,
      "hi": 11224.24,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0.004222159090909091,
      "lo": 0.004222159090909091,
      "hi": 0.004222159090909091,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=hard_negative",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 4906,
      "lo": 4906,
      "hi": 4906,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 10218.499999999996,
      "lo": 10218.499999999996,
      "hi": 10218.499999999996,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 16927.240000000005,
      "lo": 16927.240000000005,
      "hi": 16927.240000000005,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0.004225975000000002,
      "lo": 0.004225975000000002,
      "hi": 0.004225975000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=manual_edit",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 4566,
      "lo": 4566,
      "hi": 4566,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 8389.05,
      "lo": 8389.05,
      "hi": 8389.05,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 8897.009999999998,
      "lo": 8897.009999999998,
      "hi": 8897.009999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 0.004588125,
      "lo": 0.004588125,
      "hi": 0.004588125,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=EUR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 4558,
      "lo": 4558,
      "hi": 4558,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 12373.19999999998,
      "lo": 12373.19999999998,
      "hi": 12373.19999999998,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 22000.75999999975,
      "lo": 22000.75999999975,
      "hi": 22000.75999999975,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0.004323590659340656,
      "lo": 0.004323590659340656,
      "hi": 0.004323590659340656,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=IDR",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 4035,
      "lo": 4035,
      "hi": 4035,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 11347.699999999995,
      "lo": 11347.699999999995,
      "hi": 11347.699999999995,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 22252.980000000007,
      "lo": 22252.980000000007,
      "hi": 22252.980000000007,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0.004406693749999999,
      "lo": 0.004406693749999999,
      "hi": 0.004406693749999999,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=USD",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 4515,
      "lo": 4515,
      "hi": 4515,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 12590,
      "lo": 12590,
      "hi": 12590,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 26286.60000000013,
      "lo": 26286.60000000013,
      "hi": 26286.60000000013,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0.004377535714285709,
      "lo": 0.004377535714285709,
      "hi": 0.004377535714285709,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "docType=receipt",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 4137,
      "lo": 4137,
      "hi": 4137,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 6441.399999999992,
      "lo": 6441.399999999992,
      "hi": 6441.399999999992,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 9920.279999999992,
      "lo": 9920.279999999992,
      "hi": 9920.279999999992,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 0.0040964999999999994,
      "lo": 0.0040964999999999994,
      "hi": 0.0040964999999999994,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=items+total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 5065.5,
      "lo": 5065.5,
      "hi": 5065.5,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 19539.75,
      "lo": 19539.75,
      "hi": 19539.75,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 39046.29999999996,
      "lo": 39046.29999999996,
      "hi": 39046.29999999996,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0.004832416666666667,
      "lo": 0.004832416666666667,
      "hi": 0.004832416666666667,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=total",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 4175,
      "lo": 4175,
      "hi": 4175,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 12705.8,
      "lo": 12705.8,
      "hi": 12705.8,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 39363.83999999991,
      "lo": 39363.83999999991,
      "hi": 39363.83999999991,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0.004439433333333333,
      "lo": 0.004439433333333333,
      "hi": 0.004439433333333333,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=en",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 4558,
      "lo": 4558,
      "hi": 4558,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 12373.19999999998,
      "lo": 12373.19999999998,
      "hi": 12373.19999999998,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 22000.75999999975,
      "lo": 22000.75999999975,
      "hi": 22000.75999999975,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0.004323590659340656,
      "lo": 0.004323590659340656,
      "hi": 0.004323590659340656,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=id",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 4531,
      "lo": 4531,
      "hi": 4531,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 6071.65,
      "lo": 6071.65,
      "hi": 6071.65,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 6459.09,
      "lo": 6459.09,
      "hi": 6459.09,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0.004436299999999999,
      "lo": 0.004436299999999999,
      "hi": 0.004436299999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 5072.5,
      "lo": 5072.5,
      "hi": 5072.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 10985.29999999999,
      "lo": 10985.29999999999,
      "hi": 10985.29999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 14888.270000000002,
      "lo": 14888.270000000002,
      "hi": 14888.270000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0.00469655,
      "lo": 0.00469655,
      "hi": 0.00469655,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=free",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 4831,
      "lo": 4831,
      "hi": 4831,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 10025.249999999998,
      "lo": 10025.249999999998,
      "hi": 10025.249999999998,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 31711.449999999895,
      "lo": 31711.449999999895,
      "hi": 31711.449999999895,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0.004464050000000001,
      "lo": 0.004464050000000001,
      "hi": 0.004464050000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=intact",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 4456.5,
      "lo": 4456.5,
      "hi": 4456.5,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 11693.200000000008,
      "lo": 11693.200000000008,
      "hi": 11693.200000000008,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 26923.05999999988,
      "lo": 26923.05999999988,
      "hi": 26923.05999999988,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0.00431576363636363,
      "lo": 0.00431576363636363,
      "hi": 0.00431576363636363,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 4715,
      "lo": 4715,
      "hi": 4715,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 15507.8,
      "lo": 15507.8,
      "hi": 15507.8,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 17762.49999999999,
      "lo": 17762.49999999999,
      "hi": 17762.49999999999,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0.004644350000000003,
      "lo": 0.004644350000000003,
      "hi": 0.004644350000000003,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=stripped",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 4546.5,
      "lo": 4546.5,
      "hi": 4546.5,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 12577.69999999998,
      "lo": 12577.69999999998,
      "hi": 12577.69999999998,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 19226.199999999997,
      "lo": 19226.199999999997,
      "hi": 19226.199999999997,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0.004325562500000001,
      "lo": 0.004325562500000001,
      "hi": 0.004325562500000001,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=cord",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 4643,
      "lo": 4643,
      "hi": 4643,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 19633.199999999953,
      "lo": 19633.199999999953,
      "hi": 19633.199999999953,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 43429.839999999975,
      "lo": 43429.839999999975,
      "hi": 43429.839999999975,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.00445505,
      "lo": 0.00445505,
      "hi": 0.00445505,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 4804,
      "lo": 4804,
      "hi": 4804,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 17371.999999999996,
      "lo": 17371.999999999996,
      "hi": 17371.999999999996,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 19360,
      "lo": 19360,
      "hi": 19360,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 0.0052609,
      "lo": 0.0052609,
      "hi": 0.0052609,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 4906,
      "lo": 4906,
      "hi": 4906,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 10218.499999999996,
      "lo": 10218.499999999996,
      "hi": 10218.499999999996,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 16927.240000000005,
      "lo": 16927.240000000005,
      "hi": 16927.240000000005,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0.004225975000000002,
      "lo": 0.004225975000000002,
      "hi": 0.004225975000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 4050.5,
      "lo": 4050.5,
      "hi": 4050.5,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 7463.949999999998,
      "lo": 7463.949999999998,
      "hi": 7463.949999999998,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 47638.25999999987,
      "lo": 47638.25999999987,
      "hi": 47638.25999999987,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0.004293860294117648,
      "lo": 0.004293860294117648,
      "hi": 0.004293860294117648,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=expressexpense",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 4762,
      "lo": 4762,
      "hi": 4762,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 7919.249999999993,
      "lo": 7919.249999999993,
      "hi": 7919.249999999993,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 13854.130000000005,
      "lo": 13854.130000000005,
      "hi": 13854.130000000005,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0.004626975000000001,
      "lo": 0.004626975000000001,
      "hi": 0.004626975000000001,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 4505,
      "lo": 4505,
      "hi": 4505,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 8300.1,
      "lo": 8300.1,
      "hi": 8300.1,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 11351.410000000003,
      "lo": 11351.410000000003,
      "hi": 11351.410000000003,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0.004505875000000002,
      "lo": 0.004505875000000002,
      "hi": 0.004505875000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 4675,
      "lo": 4675,
      "hi": 4675,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 10671.999999999995,
      "lo": 10671.999999999995,
      "hi": 10671.999999999995,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 18571.00000000001,
      "lo": 18571.00000000001,
      "hi": 18571.00000000001,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0.004358032051282049,
      "lo": 0.004358032051282049,
      "hi": 0.004358032051282049,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=exclusive",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 4468,
      "lo": 4468,
      "hi": 4468,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 13129.499999999987,
      "lo": 13129.499999999987,
      "hi": 13129.499999999987,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 28686.539999999983,
      "lo": 28686.539999999983,
      "hi": 28686.539999999983,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0.004271136363636366,
      "lo": 0.004271136363636366,
      "hi": 0.004271136363636366,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=none",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 3835,
      "lo": 3835,
      "hi": 3835,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 5011.4,
      "lo": 5011.4,
      "hi": 5011.4,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 5077.48,
      "lo": 5077.48,
      "hi": 5077.48,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 0.00429195,
      "lo": 0.00429195,
      "hi": 0.00429195,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=crop",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 3839,
      "lo": 3839,
      "hi": 3839,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 8415.499999999993,
      "lo": 8415.499999999993,
      "hi": 8415.499999999993,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 11974.299999999996,
      "lo": 11974.299999999996,
      "hi": 11974.299999999996,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 0.0041756,
      "lo": 0.0041756,
      "hi": 0.0041756,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=jpeg",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 4587,
      "lo": 4587,
      "hi": 4587,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 19228.199999999957,
      "lo": 19228.199999999957,
      "hi": 19228.199999999957,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 42156.839999999975,
      "lo": 42156.839999999975,
      "hi": 42156.839999999975,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 0.00429755,
      "lo": 0.00429755,
      "hi": 0.00429755,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rephoto",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 4675,
      "lo": 4675,
      "hi": 4675,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 15927.69999999999,
      "lo": 15927.69999999999,
      "hi": 15927.69999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 21437.539999999994,
      "lo": 21437.539999999994,
      "hi": 21437.539999999994,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 0.00439495,
      "lo": 0.00439495,
      "hi": 0.00439495,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=resize",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 4815,
      "lo": 4815,
      "hi": 4815,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 12990.499999999993,
      "lo": 12990.499999999993,
      "hi": 12990.499999999993,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 16571.699999999997,
      "lo": 16571.699999999997,
      "hi": 16571.699999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 0.0043431500000000005,
      "lo": 0.0043431500000000005,
      "hi": 0.0043431500000000005,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rotate",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 4786,
      "lo": 4786,
      "hi": 4786,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 9183.699999999997,
      "lo": 9183.699999999997,
      "hi": 9183.699999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 10846.339999999998,
      "lo": 10846.339999999998,
      "hi": 10846.339999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 0.01633333333333333,
      "lo": 0.01633333333333333,
      "hi": 0.01633333333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 0.00444875,
      "lo": 0.00444875,
      "hi": 0.00444875,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=screenshot",
      "target": "self-gemini",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9213484012671117,
      "hi": 1,
      "n": 45,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9123783988027135,
      "hi": 1,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9057971014492755,
      "lo": 0.7876025758463708,
      "hi": 0.9614391700551873,
      "n": 46,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9626436781609196,
      "lo": 0.9107648421798313,
      "hi": 0.9848628845647558,
      "n": 116,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222221,
      "lo": 0.9251355540193794,
      "hi": 0.9900130110296227,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222221,
      "lo": 0.9251355540193794,
      "hi": 0.9900130110296227,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8620237953250198,
      "hi": 1,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8861111111111112,
      "lo": 0.81691675535672,
      "hi": 0.9313517772421702,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9172852598091198,
      "lo": 0.8657960557960558,
      "hi": 0.9541109253065775,
      "n": 98,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8864866068260312,
      "hi": 0.9999999999999999,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9313725490196078,
      "lo": 0.7958966517962754,
      "hi": 0.9792672622675235,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9818840579710144,
      "lo": 0.9296147941856543,
      "hi": 0.9955241567202318,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.96875,
      "lo": 0.9121132067546303,
      "hi": 0.989315929628799,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9756944444444443,
      "lo": 0.9220728923934436,
      "hi": 0.9927107497141845,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9097222222222223,
      "lo": 0.8355762023678344,
      "hi": 0.9523396353599038,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9275184275184275,
      "lo": 0.8797300226623312,
      "hi": 0.9652989625060482,
      "n": 81,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9761904761904762,
      "lo": 0.749095974501273,
      "hi": 0.9982270705105024,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9523809523809523,
      "lo": 0.7162271925814832,
      "hi": 0.9937297001797036,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7888888888888889,
      "lo": 0.6216216216216216,
      "hi": 0.9722499999999998,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7666666666666666,
      "lo": 0.45787114703720166,
      "hi": 0.9274446930359239,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9761904761904762,
      "lo": 0.749095974501273,
      "hi": 0.9982270705105024,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9761904761904762,
      "lo": 0.749095974501273,
      "hi": 0.9982270705105024,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9545454545454546,
      "lo": 0.8333333333333334,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9059421741627975,
      "hi": 1,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.895730002457548,
      "hi": 1,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8859649122807018,
      "lo": 0.7479073171955231,
      "hi": 0.9531517318935001,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.966360856269113,
      "lo": 0.9136238491676272,
      "hi": 0.9873452425933691,
      "n": 109,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9702380952380951,
      "lo": 0.9199927306857199,
      "hi": 0.9892960024298922,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9702380952380951,
      "lo": 0.9199927306857199,
      "hi": 0.9892960024298922,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8839285714285714,
      "lo": 0.8114943070899338,
      "hi": 0.9308996585550914,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9261430246189918,
      "lo": 0.8801274709682418,
      "hi": 0.9622657663478935,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9668386529330562,
      "hi": 1,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9833333333333332,
      "lo": 0.812152827904405,
      "hi": 0.9987595193112027,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9583333333333333,
      "lo": 0.744676139431561,
      "hi": 0.9945168200785387,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9666666666666666,
      "lo": 0.7873326750830549,
      "hi": 0.9956171774009801,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9166666666666666,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8699186991869918,
      "lo": 0.7670161290322581,
      "hi": 0.9540229885057471,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8666666666666668,
      "lo": 0.5589289935914745,
      "hi": 0.9708802865090735,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8241207763533425,
      "hi": 1,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8947368421052632,
      "lo": 0.6860591728889099,
      "hi": 0.9706414397011299,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9613526570048309,
      "lo": 0.8865112175870299,
      "hi": 0.9875331480452944,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9758454106280192,
      "lo": 0.9074744724784117,
      "hi": 0.9940267842208964,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9758454106280192,
      "lo": 0.9074744724784117,
      "hi": 0.9940267842208964,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.898550724637681,
      "lo": 0.80508114245052,
      "hi": 0.9499833618915401,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.933852140077821,
      "lo": 0.8667928443961459,
      "hi": 0.9830532279565797,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9472627418109483,
      "hi": 1,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9090909090909091,
      "lo": 0.6226415635484043,
      "hi": 0.9837678271141151,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9393939393939394,
      "lo": 0.6593196035093031,
      "hi": 0.9920090012763664,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8484848484848484,
      "lo": 0.5547884179362573,
      "hi": 0.9617825444799631,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6666666666666666,
      "lo": 0.3798489608652487,
      "hi": 0.8672067168120741,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9027777777777778,
      "lo": 0.7424242424242424,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9230769230769231,
      "lo": 0.7272727272727273,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9777777777777777,
      "lo": 0.9144909766147252,
      "hi": 0.9945062405683717,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9555555555555554,
      "lo": 0.8827435240680951,
      "hi": 0.9839747527809036,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9688888888888888,
      "lo": 0.9014262781928679,
      "hi": 0.9906593628565896,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9111111111111112,
      "lo": 0.825146388943418,
      "hi": 0.9570140072373864,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9340329835082459,
      "lo": 0.8778964397366285,
      "hi": 0.9758893617021276,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8333333333333333,
      "lo": 0.6499948702339979,
      "hi": 0.9308525151931917,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9102564102564104,
      "lo": 0.7420976017868207,
      "hi": 0.9727914879697207,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9615384615384616,
      "lo": 0.8110722109696524,
      "hi": 0.9931780150064564,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7435897435897434,
      "lo": 0.552460234772849,
      "hi": 0.872005162270097,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.916201209134296,
      "hi": 0.9999999999999999,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8992248062015505,
      "lo": 0.7742693719179959,
      "hi": 0.9586995448862176,
      "n": 43,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9646017699115044,
      "lo": 0.9125037991051361,
      "hi": 0.9861498214486251,
      "n": 113,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9715099715099714,
      "lo": 0.9232863704396193,
      "hi": 0.9897556799468059,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9715099715099714,
      "lo": 0.9232863704396193,
      "hi": 0.9897556799468059,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.886039886039886,
      "lo": 0.815798477252183,
      "hi": 0.9317374612817181,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9151251360174102,
      "lo": 0.8647706140983134,
      "hi": 0.9558757370232779,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9682107543372667,
      "hi": 1,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222222,
      "lo": 0.7174912732783316,
      "hi": 0.9979310537694965,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9166666666666666,
      "lo": 0.7894338118022329,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8668077490609515,
      "hi": 0.9999999999999999,
      "n": 25,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8602150537634408,
      "lo": 0.6987003618915625,
      "hi": 0.9422984392283383,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.924731182795699,
      "lo": 0.7785690051121716,
      "hi": 0.9772355514321893,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.967741935483871,
      "lo": 0.838058948352751,
      "hi": 0.9942827784745831,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7849462365591398,
      "lo": 0.6135338340026802,
      "hi": 0.8935249191220175,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8823529411764706,
      "lo": 0.7450189753320685,
      "hi": 0.9791666666666666,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9777777777777777,
      "lo": 0.9144909766147252,
      "hi": 0.9945062405683717,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9555555555555554,
      "lo": 0.8827435240680951,
      "hi": 0.9839747527809036,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9688888888888888,
      "lo": 0.9014262781928679,
      "hi": 0.9906593628565896,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9111111111111112,
      "lo": 0.825146388943418,
      "hi": 0.9570140072373864,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9340329835082459,
      "lo": 0.8783345894547745,
      "hi": 0.9758220335868886,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9333333333333333,
      "lo": 0.6349338104169442,
      "hi": 0.9912044297018849,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8954248366013072,
      "lo": 0.7484276729559748,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9285714285714286,
      "lo": 0.6853129557584889,
      "hi": 0.9872777847521093,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.980392156862745,
      "lo": 0.7851540058719944,
      "hi": 0.9985403350018538,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9333333333333332,
      "lo": 0.7414447301258664,
      "hi": 0.9855801328950236,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8955223880597015,
      "lo": 0.7843068213200773,
      "hi": 0.9818262987012987,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.85,
      "lo": 0.6395811352592431,
      "hi": 0.9476312541037835,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9541666666666666,
      "lo": 0.883993999656376,
      "hi": 0.9827211993488533,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9708333333333332,
      "lo": 0.9072753191609725,
      "hi": 0.9912459421930725,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9583333333333333,
      "lo": 0.8896831488976221,
      "hi": 0.984983565694811,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8666666666666666,
      "lo": 0.7751888526542121,
      "hi": 0.9245445190197346,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9300341296928327,
      "lo": 0.8655816216216217,
      "hi": 0.9760622204682977,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.9541818704644728,
      "hi": 0.9999999999999999,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222222,
      "lo": 0.7174912732783316,
      "hi": 0.9979310537694965,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9824561403508772,
      "lo": 0.8039399248192856,
      "hi": 0.9986941572352075,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9166666666666667,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8974358974358975,
      "lo": 0.7908248081841432,
      "hi": 0.9722222222222222,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222221,
      "lo": 0.8944884254780954,
      "hi": 0.9931270474988192,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.961111111111111,
      "lo": 0.8784211269356963,
      "hi": 0.9883092760888206,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.961111111111111,
      "lo": 0.8784211269356963,
      "hi": 0.9883092760888206,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9055555555555556,
      "lo": 0.8053786069519123,
      "hi": 0.9569264463106147,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9300341296928327,
      "lo": 0.8695635679883882,
      "hi": 0.9748074198191714,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9230769230769231,
      "lo": 0.7142857142857143,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.85,
      "lo": 0.6395811352592431,
      "hi": 0.9476312541037835,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.6989663547715128,
      "hi": 0.9721335187862319,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7499999999999999,
      "lo": 0.531299122381256,
      "hi": 0.8881382985923343,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9333333333333333,
      "lo": 0.6349338104169442,
      "hi": 0.9912044297018849,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8583333333333333,
      "lo": 0.689922480620155,
      "hi": 0.9833434959349593,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9540229885057471,
      "lo": 0.81174756337716,
      "hi": 0.9900844586747997,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9540229885057471,
      "lo": 0.81174756337716,
      "hi": 0.9900844586747997,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9377162629757786,
      "lo": 0.8295454545454546,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9506172839506172,
      "lo": 0.7996364310196304,
      "hi": 0.989344838472267,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9753086419753086,
      "lo": 0.835456806450102,
      "hi": 0.9967563134249131,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9135802469135802,
      "lo": 0.7503195676549773,
      "hi": 0.9738139262622437,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8641975308641976,
      "lo": 0.6897009869172017,
      "hi": 0.9479688062337841,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9108527131782945,
      "lo": 0.8333333333333334,
      "hi": 0.9676401503288444,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.903125,
      "lo": 0.83187109375,
      "hi": 0.96079296875,
      "n": 120,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.475,
      "lo": 0.325,
      "hi": 0.775,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7,
      "lo": 0.525,
      "hi": 0.875,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8,
      "lo": 0.625,
      "hi": 0.925,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5083333333333333,
      "lo": 0.3596171788908032,
      "hi": 0.6555891277559087,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.33333333333333337,
      "lo": 0.20763464534227963,
      "hi": 0.48823922172348266,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.004166666666666667,
      "lo": 0.0003098714479059343,
      "hi": 0.053459773674825115,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5714285714285714,
      "lo": 0.3259064457965074,
      "hi": 0.7861920109552588,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.40476190476190477,
      "lo": 0.19658613592958177,
      "hi": 0.6539492550680632,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8849252013808976,
      "lo": 0.8009110088224012,
      "hi": 0.9558879938626774,
      "n": 112,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.48484848484848486,
      "lo": 0.3333333333333333,
      "hi": 0.7272727272727273,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6060606060606061,
      "lo": 0.42424242424242425,
      "hi": 0.8181818181818182,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.696969696969697,
      "lo": 0.5151515151515151,
      "hi": 0.8787878787878788,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.49494949494949503,
      "lo": 0.3340291630548903,
      "hi": 0.6569230591425084,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.008438818565400843,
      "lo": 0.0011067800440412306,
      "hi": 0.061359432257072605,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.35353535353535354,
      "lo": 0.21368165803590483,
      "hi": 0.5239327856886518,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.004219409282700422,
      "lo": 0.00031379493234937397,
      "hi": 0.05410491704516887,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8844444444444445,
      "lo": 0.7722222222222223,
      "hi": 0.9766666666666667,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6,
      "lo": 0.4,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7333333333333333,
      "lo": 0.4666666666666667,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6,
      "lo": 0.357468301206246,
      "hi": 0.8017550385865841,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.01111111111111111,
      "lo": 0.00145765367836867,
      "hi": 0.07959926841611918,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.4,
      "lo": 0.1982449614134158,
      "hi": 0.6425316987937539,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8715277777777778,
      "lo": 0.663107638888889,
      "hi": 1,
      "n": 26,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8991554054054054,
      "lo": 0.8248268581081081,
      "hi": 0.9579434121621622,
      "n": 117,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5135135135135135,
      "lo": 0.35135135135135137,
      "hi": 0.7837837837837838,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7027027027027027,
      "lo": 0.5405405405405406,
      "hi": 0.8648648648648649,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7837837837837838,
      "lo": 0.6209459459459471,
      "hi": 0.918918918918919,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5315315315315314,
      "lo": 0.37549799276776696,
      "hi": 0.6816334956929498,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.35135135135135137,
      "lo": 0.2182560713392713,
      "hi": 0.5124097687744917,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.004166666666666667,
      "lo": 0.0003098714479059343,
      "hi": 0.053459773674825115,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9722222222222223,
      "lo": 0.7174912732783317,
      "hi": 0.9979310537694964,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9100877192982456,
      "lo": 0.7565241228070176,
      "hi": 1,
      "n": 31,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.47222222222222215,
      "lo": 0.23302775994579422,
      "hi": 0.7248885737572746,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.3055555555555555,
      "lo": 0.1210211581599055,
      "hi": 0.584393177761577,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8844444444444445,
      "lo": 0.7616527777777778,
      "hi": 0.9744444444444444,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6,
      "lo": 0.3333333333333333,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7333333333333333,
      "lo": 0.4666666666666667,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6,
      "lo": 0.357468301206246,
      "hi": 0.8017550385865841,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.01111111111111111,
      "lo": 0.00145765367836867,
      "hi": 0.07959926841611918,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.4,
      "lo": 0.1982449614134158,
      "hi": 0.6425316987937539,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.19999999999999998,
      "lo": 0.056682151454375246,
      "hi": 0.5098375284633582,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.13333333333333333,
      "lo": 0.029119713490926458,
      "hi": 0.4410710064085256,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6333333333333333,
      "lo": 0.339777586667533,
      "hi": 0.8528803333690298,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.3,
      "lo": 0.10779126740630099,
      "hi": 0.6032218525388546,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.01111111111111111,
      "lo": 0.00145765367836867,
      "hi": 0.07959926841611918,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5333333333333333,
      "lo": 0.26109856576627405,
      "hi": 0.7870659142428666,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0.36666666666666664,
      "lo": 0.1471196666309701,
      "hi": 0.6602224133324669,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5,
      "lo": 0.236593090512564,
      "hi": 0.7634069094874361,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.3333333333333333,
      "lo": 0.12697541213017843,
      "hi": 0.632202187824118,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.33333333333333337,
      "lo": 0.1269754121301785,
      "hi": 0.632202187824118,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.1,
      "lo": 0.017876213095072896,
      "hi": 0.4041500267952385,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.012345679012345678,
      "lo": 0.0009186195639853539,
      "hi": 0.14525233536944054,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.012345679012345678,
      "lo": 0.0009186195639853539,
      "hi": 0.14525233536944054,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5,
      "lo": 0.2992980081982123,
      "hi": 0.7007019918017877,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.35,
      "lo": 0.18119182410108206,
      "hi": 0.5671457233147638,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5166666666666666,
      "lo": 0.3133728128430116,
      "hi": 0.7145896818885611,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.31666666666666665,
      "lo": 0.15716056329227926,
      "hi": 0.5352519946604213,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.004166666666666667,
      "lo": 0.0003098714479059343,
      "hi": 0.053459773674825115,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9503125,
      "lo": 0.8959296875,
      "hi": 0.9903203125,
      "n": 100,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.35,
      "lo": 0.15,
      "hi": 0.8,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.75,
      "lo": 0.5,
      "hi": 0.95,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.65,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.41666666666666663,
      "lo": 0.23174394158041686,
      "hi": 0.6284435847617196,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.21666666666666665,
      "lo": 0.09076519766966104,
      "hi": 0.43387239189360327,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.798125,
      "lo": 0.578109375,
      "hi": 0.9591718749999943,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6666666666666667,
      "lo": 0.3677978121758822,
      "hi": 0.8730245878698215,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5333333333333334,
      "lo": 0.26109856576627416,
      "hi": 0.7870659142428669,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.91375,
      "lo": 0.822484375,
      "hi": 0.98625,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5,
      "lo": 0.2,
      "hi": 0.8,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6,
      "lo": 0.3,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5333333333333333,
      "lo": 0.26109856576627405,
      "hi": 0.7870659142428666,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.36666666666666664,
      "lo": 0.1471196666309701,
      "hi": 0.6602224133324669,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.8509615384615384,
      "lo": 0.7492728365384616,
      "hi": 0.9331730769230769,
      "n": 106,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.19230769230769232,
      "lo": 0.07692307692307693,
      "hi": 0.5769230769230769,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5384615384615384,
      "lo": 0.3076923076923077,
      "hi": 0.7692307692307693,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6923076923076923,
      "lo": 0.46153846153846156,
      "hi": 0.8461538461538461,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.2692307692307692,
      "lo": 0.13704448322018375,
      "hi": 0.4608304037917617,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.15384615384615383,
      "lo": 0.06150033723578355,
      "hi": 0.3353119932821347,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 94,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9523809523809523,
      "lo": 0.7162271925814832,
      "hi": 0.9937297001797036,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.6666666666666666,
      "lo": 0.40911830209688393,
      "hi": 0.8524447636572373,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "decision_consistency",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9,
      "lo": 0.833318258891048,
      "hi": 0.9418664079739892,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "total_consistency",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.9833333333333333,
      "lo": 0.941264037027841,
      "hi": 0.9954174354340789,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "calibration_error",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.10722222222222225,
      "lo": 0.08497152777777776,
      "hi": 0.13911111111111113,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 631.9166666666667,
      "lo": 416.66666666666674,
      "hi": 883.9583333333334,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 736.7916666666666,
      "lo": 566.6666666666666,
      "hi": 935.375,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:claude-haiku-5.5",
      "value": 190.69166666666663,
      "lo": 44.166666666666664,
      "hi": 407.1458333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:claude-haiku-5.5",
      "value": 137.42916666666665,
      "lo": 57.49999999999999,
      "hi": 287.2875,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:claude-haiku-5.5",
      "value": 2592.916666666667,
      "lo": 1875.0000000000014,
      "hi": 3336.458333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:claude-haiku-5.5",
      "value": 3400.6250000000005,
      "lo": 2750.0000000000005,
      "hi": 4009.5833333333344,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_32usd",
      "target": "diy:claude-haiku-5.5",
      "value": 297.58333333333337,
      "lo": 141.33333333333334,
      "hi": 515.9583333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_32usd",
      "target": "diy:claude-haiku-5.5",
      "value": 283.4583333333333,
      "lo": 181.33333333333331,
      "hi": 447.70833333333337,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_182usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1035.0833333333335,
      "lo": 728,
      "hi": 1351.7677083333326,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_182usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1283.4583333333333,
      "lo": 1034.9895833333337,
      "hi": 1554.4999999999998,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_8usd",
      "target": "diy:claude-haiku-5.5",
      "value": 557.6666666666667,
      "lo": 399.66666666666686,
      "hi": 731.9999999999999,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_8usd",
      "target": "diy:claude-haiku-5.5",
      "value": 699.6666666666666,
      "lo": 558.3333333333334,
      "hi": 824.6666666666665,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_73usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1093.916666666667,
      "lo": 441.4583333333337,
      "hi": 2038.958333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_73usd",
      "target": "diy:claude-haiku-5.5",
      "value": 967.7916666666666,
      "lo": 575,
      "hi": 1670.4583333333312,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:claude-haiku-5.5",
      "value": 100,
      "lo": 100,
      "hi": 100,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:claude-haiku-5.5",
      "value": 5000,
      "lo": 5000,
      "hi": 5000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_32usd",
      "target": "diy:claude-haiku-5.5",
      "value": 320,
      "lo": 320,
      "hi": 320,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_182usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1820,
      "lo": 1820,
      "hi": 1820,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_8usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_73usd",
      "target": "diy:claude-haiku-5.5",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "evidence",
      "metric": "edited_field_named",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.5833333333333333,
      "lo": 0.3715564152382803,
      "hi": 0.7682560584195831,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 7212.5,
      "lo": 7212.5,
      "hi": 7212.5,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 11029.349999999997,
      "lo": 11029.349999999997,
      "hi": 11029.349999999997,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 16098.109999999999,
      "lo": 16098.109999999999,
      "hi": 16098.109999999999,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008123842777777787,
      "lo": 0.0008123842777777787,
      "hi": 0.0008123842777777787,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008123842777777787,
      "lo": 0.0008123842777777787,
      "hi": 0.0008123842777777787,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "all",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 7041,
      "lo": 7041,
      "hi": 7041,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 10963.15,
      "lo": 10963.15,
      "hi": 10963.15,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 16267.620000000006,
      "lo": 16267.620000000006,
      "hi": 16267.620000000006,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007939008465608472,
      "lo": 0.0007939008465608472,
      "hi": 0.0007939008465608472,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007939008465608472,
      "lo": 0.0007939008465608472,
      "hi": 0.0007939008465608472,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 8467.5,
      "lo": 8467.5,
      "hi": 8467.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 10821.149999999998,
      "lo": 10821.149999999998,
      "hi": 10821.149999999998,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 14057.089999999997,
      "lo": 14057.089999999997,
      "hi": 14057.089999999997,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009240961904761902,
      "lo": 0.0009240961904761902,
      "hi": 0.0009240961904761902,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009240961904761902,
      "lo": 0.0009240961904761902,
      "hi": 0.0009240961904761902,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 7842,
      "lo": 7842,
      "hi": 7842,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 12312.699999999997,
      "lo": 12312.699999999997,
      "hi": 12312.699999999997,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 15066.540000000003,
      "lo": 15066.540000000003,
      "hi": 15066.540000000003,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008888788333333334,
      "lo": 0.0008888788333333334,
      "hi": 0.0008888788333333334,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008888788333333334,
      "lo": 0.0008888788333333334,
      "hi": 0.0008888788333333334,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 6987.5,
      "lo": 6987.5,
      "hi": 6987.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 9462.05,
      "lo": 9462.05,
      "hi": 9462.05,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 10045.949999999997,
      "lo": 10045.949999999997,
      "hi": 10045.949999999997,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007670184523809522,
      "lo": 0.0007670184523809522,
      "hi": 0.0007670184523809522,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007670184523809522,
      "lo": 0.0007670184523809522,
      "hi": 0.0007670184523809522,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 8020,
      "lo": 8020,
      "hi": 8020,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 9938.449999999999,
      "lo": 9938.449999999999,
      "hi": 9938.449999999999,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 10266.89,
      "lo": 10266.89,
      "hi": 10266.89,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008739847222222221,
      "lo": 0.0008739847222222221,
      "hi": 0.0008739847222222221,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008739847222222221,
      "lo": 0.0008739847222222221,
      "hi": 0.0008739847222222221,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 7141.5,
      "lo": 7141.5,
      "hi": 7141.5,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 10930.599999999999,
      "lo": 10930.599999999999,
      "hi": 10930.599999999999,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 16165.730000000001,
      "lo": 16165.730000000001,
      "hi": 16165.730000000001,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007970934681372558,
      "lo": 0.0007970934681372558,
      "hi": 0.0007970934681372558,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007970934681372558,
      "lo": 0.0007970934681372558,
      "hi": 0.0007970934681372558,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=photo",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 9079.5,
      "lo": 9079.5,
      "hi": 9079.5,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 14073.649999999998,
      "lo": 14073.649999999998,
      "hi": 14073.649999999998,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 14418.539999999999,
      "lo": 14418.539999999999,
      "hi": 14418.539999999999,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010761714583333332,
      "lo": 0.0010761714583333332,
      "hi": 0.0010761714583333332,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010761714583333332,
      "lo": 0.0010761714583333332,
      "hi": 0.0010761714583333332,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=scan",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 6704,
      "lo": 6704,
      "hi": 6704,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 8721.199999999999,
      "lo": 8721.199999999999,
      "hi": 8721.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 9339.439999999999,
      "lo": 9339.439999999999,
      "hi": 9339.439999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007593733333333334,
      "lo": 0.0007593733333333334,
      "hi": 0.0007593733333333334,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007593733333333334,
      "lo": 0.0007593733333333334,
      "hi": 0.0007593733333333334,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 8639,
      "lo": 8639,
      "hi": 8639,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 13358.15,
      "lo": 13358.15,
      "hi": 13358.15,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 14309.82,
      "lo": 14309.82,
      "hi": 14309.82,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010324241666666665,
      "lo": 0.0010324241666666665,
      "hi": 0.0010324241666666665,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010324241666666665,
      "lo": 0.0010324241666666665,
      "hi": 0.0010324241666666665,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_full",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 7098,
      "lo": 7098,
      "hi": 7098,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 9793.1,
      "lo": 9793.1,
      "hi": 9793.1,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 10208.93,
      "lo": 10208.93,
      "hi": 10208.93,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008142466666666664,
      "lo": 0.0008142466666666664,
      "hi": 0.0008142466666666664,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008142466666666664,
      "lo": 0.0008142466666666664,
      "hi": 0.0008142466666666664,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 7257,
      "lo": 7257,
      "hi": 7257,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 9831,
      "lo": 9831,
      "hi": 9831,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 12325.49,
      "lo": 12325.49,
      "hi": 12325.49,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008000483888888888,
      "lo": 0.0008000483888888888,
      "hi": 0.0008000483888888888,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008000483888888888,
      "lo": 0.0008000483888888888,
      "hi": 0.0008000483888888888,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=duplicate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 6693,
      "lo": 6693,
      "hi": 6693,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 11739.699999999999,
      "lo": 11739.699999999999,
      "hi": 11739.699999999999,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 16665.94,
      "lo": 16665.94,
      "hi": 16665.94,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839824637681162,
      "lo": 0.0007839824637681162,
      "hi": 0.0007839824637681162,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839824637681162,
      "lo": 0.0007839824637681162,
      "hi": 0.0007839824637681162,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=genuine",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 6476,
      "lo": 6476,
      "hi": 6476,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 8931.4,
      "lo": 8931.4,
      "hi": 8931.4,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 9191.56,
      "lo": 9191.56,
      "hi": 9191.56,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.00064825,
      "lo": 0.00064825,
      "hi": 0.00064825,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0.00064825,
      "lo": 0.00064825,
      "hi": 0.00064825,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=hard_negative",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 7247,
      "lo": 7247,
      "hi": 7247,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 9593.15,
      "lo": 9593.15,
      "hi": 9593.15,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 10118.550000000001,
      "lo": 10118.550000000001,
      "hi": 10118.550000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839699999999999,
      "lo": 0.0007839699999999999,
      "hi": 0.0007839699999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839699999999999,
      "lo": 0.0007839699999999999,
      "hi": 0.0007839699999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=manual_edit",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 9420,
      "lo": 9420,
      "hi": 9420,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 13594.649999999998,
      "lo": 13594.649999999998,
      "hi": 13594.649999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 14309.329999999998,
      "lo": 14309.329999999998,
      "hi": 14309.329999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010680916666666665,
      "lo": 0.0010680916666666665,
      "hi": 0.0010680916666666665,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010680916666666665,
      "lo": 0.0010680916666666665,
      "hi": 0.0010680916666666665,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=EUR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 6782,
      "lo": 6782,
      "hi": 6782,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 10273,
      "lo": 10273,
      "hi": 10273,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 16287.319999999989,
      "lo": 16287.319999999989,
      "hi": 16287.319999999989,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007669015934065939,
      "lo": 0.0007669015934065939,
      "hi": 0.0007669015934065939,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007669015934065939,
      "lo": 0.0007669015934065939,
      "hi": 0.0007669015934065939,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=IDR",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 7407,
      "lo": 7407,
      "hi": 7407,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 10356.25,
      "lo": 10356.25,
      "hi": 10356.25,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 12173.560000000001,
      "lo": 12173.560000000001,
      "hi": 12173.560000000001,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008062902083333331,
      "lo": 0.0008062902083333331,
      "hi": 0.0008062902083333331,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008062902083333331,
      "lo": 0.0008062902083333331,
      "hi": 0.0008062902083333331,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=USD",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 7205,
      "lo": 7205,
      "hi": 7205,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 11055,
      "lo": 11055,
      "hi": 11055,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 16112.600000000004,
      "lo": 16112.600000000004,
      "hi": 16112.600000000004,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008102036507936519,
      "lo": 0.0008102036507936519,
      "hi": 0.0008102036507936519,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008102036507936519,
      "lo": 0.0008102036507936519,
      "hi": 0.0008102036507936519,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "docType=receipt",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 6653,
      "lo": 6653,
      "hi": 6653,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 8442.5,
      "lo": 8442.5,
      "hi": 8442.5,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 8816.5,
      "lo": 8816.5,
      "hi": 8816.5,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0006949183333333333,
      "lo": 0.0006949183333333333,
      "hi": 0.0006949183333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0006949183333333333,
      "lo": 0.0006949183333333333,
      "hi": 0.0006949183333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=items+total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 8014,
      "lo": 8014,
      "hi": 8014,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 9973,
      "lo": 9973,
      "hi": 9973,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 10329.75,
      "lo": 10329.75,
      "hi": 10329.75,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008753302777777776,
      "lo": 0.0008753302777777776,
      "hi": 0.0008753302777777776,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008753302777777776,
      "lo": 0.0008753302777777776,
      "hi": 0.0008753302777777776,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=total",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 7658,
      "lo": 7658,
      "hi": 7658,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 10599.9,
      "lo": 10599.9,
      "hi": 10599.9,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 12706.38,
      "lo": 12706.38,
      "hi": 12706.38,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008321611851851852,
      "lo": 0.0008321611851851852,
      "hi": 0.0008321611851851852,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008321611851851852,
      "lo": 0.0008321611851851852,
      "hi": 0.0008321611851851852,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=en",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 6782,
      "lo": 6782,
      "hi": 6782,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 10273,
      "lo": 10273,
      "hi": 10273,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 16287.319999999989,
      "lo": 16287.319999999989,
      "hi": 16287.319999999989,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007669015934065939,
      "lo": 0.0007669015934065939,
      "hi": 0.0007669015934065939,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007669015934065939,
      "lo": 0.0007669015934065939,
      "hi": 0.0007669015934065939,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=id",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 8473.5,
      "lo": 8473.5,
      "hi": 8473.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 11134.249999999998,
      "lo": 11134.249999999998,
      "hi": 11134.249999999998,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 11552.44,
      "lo": 11552.44,
      "hi": 11552.44,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009639491666666664,
      "lo": 0.0009639491666666664,
      "hi": 0.0009639491666666664,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009639491666666664,
      "lo": 0.0009639491666666664,
      "hi": 0.0009639491666666664,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 9316.5,
      "lo": 9316.5,
      "hi": 9316.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 13848.949999999997,
      "lo": 13848.949999999997,
      "hi": 13848.949999999997,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 14400.42,
      "lo": 14400.42,
      "hi": 14400.42,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0011008991666666668,
      "lo": 0.0011008991666666668,
      "hi": 0.0011008991666666668,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0011008991666666668,
      "lo": 0.0011008991666666668,
      "hi": 0.0011008991666666668,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=free",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 8173,
      "lo": 8173,
      "hi": 8173,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 11674.65,
      "lo": 11674.65,
      "hi": 11674.65,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 13867.909999999996,
      "lo": 13867.909999999996,
      "hi": 13867.909999999996,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009242604999999998,
      "lo": 0.0009242604999999998,
      "hi": 0.0009242604999999998,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009242604999999998,
      "lo": 0.0009242604999999998,
      "hi": 0.0009242604999999998,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=intact",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 6890,
      "lo": 6890,
      "hi": 6890,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 10758.000000000002,
      "lo": 10758.000000000002,
      "hi": 10758.000000000002,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 16456.73999999999,
      "lo": 16456.73999999999,
      "hi": 16456.73999999999,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007747908333333342,
      "lo": 0.0007747908333333342,
      "hi": 0.0007747908333333342,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007747908333333342,
      "lo": 0.0007747908333333342,
      "hi": 0.0007747908333333342,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 7992.5,
      "lo": 7992.5,
      "hi": 7992.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 10970.199999999997,
      "lo": 10970.199999999997,
      "hi": 10970.199999999997,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 13695.119999999997,
      "lo": 13695.119999999997,
      "hi": 13695.119999999997,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009072720000000001,
      "lo": 0.0009072720000000001,
      "hi": 0.0009072720000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009072720000000001,
      "lo": 0.0009072720000000001,
      "hi": 0.0009072720000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=stripped",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 6743,
      "lo": 6743,
      "hi": 6743,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 10626.999999999998,
      "lo": 10626.999999999998,
      "hi": 10626.999999999998,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 16464.619999999995,
      "lo": 16464.619999999995,
      "hi": 16464.619999999995,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007642723245614037,
      "lo": 0.0007642723245614037,
      "hi": 0.0007642723245614037,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007642723245614037,
      "lo": 0.0007642723245614037,
      "hi": 0.0007642723245614037,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=cord",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 6981,
      "lo": 6981,
      "hi": 6981,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 10010.9,
      "lo": 10010.9,
      "hi": 10010.9,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 10281.38,
      "lo": 10281.38,
      "hi": 10281.38,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.000831939,
      "lo": 0.000831939,
      "hi": 0.000831939,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.000831939,
      "lo": 0.000831939,
      "hi": 0.000831939,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 7453,
      "lo": 7453,
      "hi": 7453,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 9424,
      "lo": 9424,
      "hi": 9424,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 9648,
      "lo": 9648,
      "hi": 9648,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007965543333333331,
      "lo": 0.0007965543333333331,
      "hi": 0.0007965543333333331,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007965543333333331,
      "lo": 0.0007965543333333331,
      "hi": 0.0007965543333333331,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 7247,
      "lo": 7247,
      "hi": 7247,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 9593.15,
      "lo": 9593.15,
      "hi": 9593.15,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 10118.550000000001,
      "lo": 10118.550000000001,
      "hi": 10118.550000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839699999999999,
      "lo": 0.0007839699999999999,
      "hi": 0.0007839699999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007839699999999999,
      "lo": 0.0007839699999999999,
      "hi": 0.0007839699999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 7319,
      "lo": 7319,
      "hi": 7319,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 10978.899999999996,
      "lo": 10978.899999999996,
      "hi": 10978.899999999996,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 12819.569999999998,
      "lo": 12819.569999999998,
      "hi": 12819.569999999998,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007983027941176469,
      "lo": 0.0007983027941176469,
      "hi": 0.0007983027941176469,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007983027941176469,
      "lo": 0.0007983027941176469,
      "hi": 0.0007983027941176469,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=expressexpense",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 9321,
      "lo": 9321,
      "hi": 9321,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 12753.249999999996,
      "lo": 12753.249999999996,
      "hi": 12753.249999999996,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 13944.720000000001,
      "lo": 13944.720000000001,
      "hi": 13944.720000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.00105509,
      "lo": 0.00105509,
      "hi": 0.00105509,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0.00105509,
      "lo": 0.00105509,
      "hi": 0.00105509,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 8546,
      "lo": 8546,
      "hi": 8546,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 12472.649999999994,
      "lo": 12472.649999999994,
      "hi": 12472.649999999994,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 14183.210000000001,
      "lo": 14183.210000000001,
      "hi": 14183.210000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010097583333333333,
      "lo": 0.0010097583333333333,
      "hi": 0.0010097583333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0010097583333333333,
      "lo": 0.0010097583333333333,
      "hi": 0.0010097583333333333,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 7006,
      "lo": 7006,
      "hi": 7006,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 11090.999999999993,
      "lo": 11090.999999999993,
      "hi": 11090.999999999993,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 19287.560000000012,
      "lo": 19287.560000000012,
      "hi": 19287.560000000012,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0.000838345085470085,
      "lo": 0.000838345085470085,
      "hi": 0.000838345085470085,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0.000838345085470085,
      "lo": 0.000838345085470085,
      "hi": 0.000838345085470085,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 6366,
      "lo": 6366,
      "hi": 6366,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 8996.199999999999,
      "lo": 8996.199999999999,
      "hi": 8996.199999999999,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 10141.759999999978,
      "lo": 10141.759999999978,
      "hi": 10141.759999999978,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0006575958585858583,
      "lo": 0.0006575958585858583,
      "hi": 0.0006575958585858583,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0006575958585858583,
      "lo": 0.0006575958585858583,
      "hi": 0.0006575958585858583,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=none",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 7508,
      "lo": 7508,
      "hi": 7508,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 10003.5,
      "lo": 10003.5,
      "hi": 10003.5,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 10207.9,
      "lo": 10207.9,
      "hi": 10207.9,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008399033333333331,
      "lo": 0.0008399033333333331,
      "hi": 0.0008399033333333331,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0008399033333333331,
      "lo": 0.0008399033333333331,
      "hi": 0.0008399033333333331,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=crop",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 6596,
      "lo": 6596,
      "hi": 6596,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 9351.5,
      "lo": 9351.5,
      "hi": 9351.5,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 9483.1,
      "lo": 9483.1,
      "hi": 9483.1,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007076646666666669,
      "lo": 0.0007076646666666669,
      "hi": 0.0007076646666666669,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007076646666666669,
      "lo": 0.0007076646666666669,
      "hi": 0.0007076646666666669,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=jpeg",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 7258,
      "lo": 7258,
      "hi": 7258,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 8785.399999999998,
      "lo": 8785.399999999998,
      "hi": 8785.399999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 9506.68,
      "lo": 9506.68,
      "hi": 9506.68,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007719966666666667,
      "lo": 0.0007719966666666667,
      "hi": 0.0007719966666666667,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007719966666666667,
      "lo": 0.0007719966666666667,
      "hi": 0.0007719966666666667,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rephoto",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 7462,
      "lo": 7462,
      "hi": 7462,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 9505.1,
      "lo": 9505.1,
      "hi": 9505.1,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 9531.42,
      "lo": 9531.42,
      "hi": 9531.42,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007950303333333333,
      "lo": 0.0007950303333333333,
      "hi": 0.0007950303333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007950303333333333,
      "lo": 0.0007950303333333333,
      "hi": 0.0007950303333333333,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=resize",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 7494,
      "lo": 7494,
      "hi": 7494,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 12431.699999999999,
      "lo": 12431.699999999999,
      "hi": 12431.699999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 12744.74,
      "lo": 12744.74,
      "hi": 12744.74,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009263219999999999,
      "lo": 0.0009263219999999999,
      "hi": 0.0009263219999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0009263219999999999,
      "lo": 0.0009263219999999999,
      "hi": 0.0009263219999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rotate",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 6704,
      "lo": 6704,
      "hi": 6704,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 8721.199999999999,
      "lo": 8721.199999999999,
      "hi": 8721.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 9339.439999999999,
      "lo": 9339.439999999999,
      "hi": 9339.439999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007593733333333334,
      "lo": 0.0007593733333333334,
      "hi": 0.0007593733333333334,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0.0007593733333333334,
      "lo": 0.0007593733333333334,
      "hi": 0.0007593733333333334,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=screenshot",
      "target": "diy:claude-haiku-5.5",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9213484012671117,
      "hi": 1,
      "n": 45,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9123783988027135,
      "hi": 1,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9229264369144198,
      "hi": 1,
      "n": 46,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9339080459770116,
      "lo": 0.8733948490818166,
      "hi": 0.9666038263121752,
      "n": 116,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9916666666666667,
      "lo": 0.9543037087549505,
      "hi": 0.998527444266658,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9916666666666667,
      "lo": 0.9543037087549505,
      "hi": 0.998527444266658,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8620237953250198,
      "hi": 1,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9277777777777778,
      "lo": 0.8670263635291507,
      "hi": 0.9619905718681808,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9670912951167728,
      "lo": 0.9244858696470234,
      "hi": 0.9933115942028985,
      "n": 98,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8864866068260312,
      "hi": 0.9999999999999999,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8984854458466761,
      "hi": 0.9999999999999999,
      "n": 34,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9673913043478263,
      "lo": 0.9084664004700646,
      "hi": 0.9888488225890285,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9895833333333334,
      "lo": 0.9433338237728424,
      "hi": 0.9981588297831838,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9583333333333335,
      "lo": 0.8977180853334191,
      "hi": 0.9836792924637118,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.985239852398524,
      "lo": 0.9702842377260982,
      "hi": 0.9973014866483477,
      "n": 81,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9615244121423243,
      "hi": 1,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8809523809523808,
      "lo": 0.6277317275844682,
      "hi": 0.970126708424952,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8809523809523808,
      "lo": 0.6277317275844682,
      "hi": 0.970126708424952,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7888888888888889,
      "lo": 0.5405405405405406,
      "hi": 1,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7,
      "lo": 0.39677814746114537,
      "hi": 0.8922087325936989,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9761904761904763,
      "lo": 0.7490959745012732,
      "hi": 0.9982270705105026,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9761904761904763,
      "lo": 0.7490959745012732,
      "hi": 0.9982270705105026,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9059421741627975,
      "hi": 1,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.895730002457548,
      "hi": 1,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9480122324159023,
      "lo": 0.889052181178457,
      "hi": 0.9764689463165334,
      "n": 109,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9910714285714286,
      "lo": 0.9511515316911152,
      "hi": 0.9984221452967079,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9910714285714286,
      "lo": 0.9511515316911152,
      "hi": 0.9984221452967079,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9404761904761906,
      "lo": 0.880374853071378,
      "hi": 0.9713639602267907,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9859154929577465,
      "lo": 0.972132497075952,
      "hi": 0.9974937343358395,
      "n": 92,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9668386529330562,
      "hi": 1,
      "n": 112,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0.875,
      "lo": 0.639771727342413,
      "hi": 0.9650225122567595,
      "n": 16,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.6989663547715128,
      "hi": 0.9721335187862319,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9065040650406504,
      "lo": 0.771083323191757,
      "hi": 0.991869918699187,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9333333333333332,
      "lo": 0.6349338104169441,
      "hi": 0.991204429701885,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8241207763533425,
      "hi": 1,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9613526570048311,
      "lo": 0.8865112175870302,
      "hi": 0.9875331480452945,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9855072463768116,
      "lo": 0.9223687753885529,
      "hi": 0.9974370753554117,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9855072463768116,
      "lo": 0.9223687753885529,
      "hi": 0.9974370753554117,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9468599033816426,
      "lo": 0.866646548245572,
      "hi": 0.9799409263197684,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9883040935672515,
      "lo": 0.9709259825921392,
      "hi": 1,
      "n": 50,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9472627418109483,
      "hi": 1,
      "n": 69,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8181818181818182,
      "lo": 0.523019438039156,
      "hi": 0.9486323102539149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8181818181818182,
      "lo": 0.523019438039156,
      "hi": 0.9486323102539149,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9733333333333335,
      "lo": 0.9078851181224088,
      "hi": 0.9926563109938686,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9600000000000002,
      "lo": 0.8888701291759629,
      "hi": 0.9863039357398562,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.990990990990991,
      "lo": 0.9771663315319844,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8205128205128205,
      "lo": 0.6355161760510599,
      "hi": 0.9229909253212378,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9615384615384616,
      "lo": 0.8110722109696524,
      "hi": 0.9931780150064564,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8205128205128205,
      "lo": 0.6355161760510599,
      "hi": 0.9229909253212378,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8712710781407846,
      "hi": 1,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.916201209134296,
      "hi": 0.9999999999999999,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.908190131774416,
      "hi": 0.9999999999999999,
      "n": 38,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9179901967742088,
      "hi": 1,
      "n": 43,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9410029498525074,
      "lo": 0.8813877250402206,
      "hi": 0.9716199973266831,
      "n": 113,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9914529914529915,
      "lo": 0.9531704967258107,
      "hi": 0.9984896464262035,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9914529914529915,
      "lo": 0.9531704967258107,
      "hi": 0.9984896464262035,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8453609810751531,
      "hi": 1,
      "n": 21,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9344729344729346,
      "lo": 0.8744302221012387,
      "hi": 0.966892513149093,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9662309368191722,
      "lo": 0.9255644393679802,
      "hi": 0.993243811993812,
      "n": 96,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9682107543372667,
      "hi": 1,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9444444444444445,
      "lo": 0.6806599518783091,
      "hi": 0.9926787088725882,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9444444444444445,
      "lo": 0.6806599518783091,
      "hi": 0.9926787088725882,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666667,
      "lo": 0.9047619047619048,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8668077490609515,
      "hi": 0.9999999999999999,
      "n": 25,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8494623655913978,
      "lo": 0.686097986159129,
      "hi": 0.9357665223900284,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.967741935483871,
      "lo": 0.838058948352751,
      "hi": 0.9942827784745831,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8494623655913978,
      "lo": 0.686097986159129,
      "hi": 0.9357665223900284,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9411764705882353,
      "lo": 0.8787878787878788,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.889744604539564,
      "hi": 1,
      "n": 31,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9733333333333335,
      "lo": 0.9078851181224088,
      "hi": 0.9926563109938686,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9866666666666667,
      "lo": 0.9282663444557845,
      "hi": 0.9976424488609515,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9600000000000002,
      "lo": 0.8888701291759629,
      "hi": 0.9863039357398562,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.990990990990991,
      "lo": 0.977098361229778,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9512761575171942,
      "hi": 1,
      "n": 75,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8888888888888888,
      "lo": 0.6792452830188679,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7846891972623643,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9019607843137255,
      "lo": 0.6802647820236702,
      "hi": 0.9754794623809784,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7719046276458016,
      "hi": 0.9999999999999999,
      "n": 13,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9166666666666667,
      "lo": 0.7198571898005176,
      "hi": 0.9792051784887997,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8955223880597015,
      "lo": 0.7303494072047424,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9416666666666668,
      "lo": 0.8673348107018559,
      "hi": 0.9755258415417618,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9875,
      "lo": 0.9325372992233133,
      "hi": 0.9977900244795476,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9875,
      "lo": 0.9325372992233133,
      "hi": 0.9977900244795476,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9291666666666668,
      "lo": 0.8511744795251084,
      "hi": 0.9678316259568978,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9897435897435898,
      "lo": 0.9748371803467101,
      "hi": 1,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9541818704644728,
      "hi": 0.9999999999999999,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7575059933447592,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9298245614035087,
      "lo": 0.7302604185822232,
      "hi": 0.9848135817935975,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7411670330319684,
      "hi": 1,
      "n": 11,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9333333333333332,
      "lo": 0.7414447301258664,
      "hi": 0.9855801328950236,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9743589743589743,
      "lo": 0.9319727891156463,
      "hi": 1,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9722222222222222,
      "lo": 0.8944884254780956,
      "hi": 0.9931270474988194,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9833333333333333,
      "lo": 0.9114487027240993,
      "hi": 0.9970518402052136,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9555555555555556,
      "lo": 0.870638772907186,
      "hi": 0.9856490951411321,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9897435897435898,
      "lo": 0.9734006997383093,
      "hi": 1,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.93982814785791,
      "hi": 0.9999999999999998,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666666,
      "lo": 0.6767863445664513,
      "hi": 0.9975163755615184,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8318207759473837,
      "hi": 1,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8156818649911483,
      "hi": 1,
      "n": 17,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.85,
      "lo": 0.6395811352592431,
      "hi": 0.9476312541037835,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.95,
      "lo": 0.763868806553258,
      "hi": 0.9911185511992047,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.85,
      "lo": 0.6395811352592431,
      "hi": 0.9476312541037835,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.merchant",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.date",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.currency",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.receipt_number",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8416666666666667,
      "lo": 0.6046511627906976,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9895833333333334,
      "lo": 0.9680851063829787,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8830302015002593,
      "hi": 1,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.subtotal",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9382716049382716,
      "lo": 0.7827176790411627,
      "hi": 0.9846476652591756,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.tax",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9629629629629629,
      "lo": 0.8171653406642414,
      "hi": 0.9934318540192149,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "field_accuracy.total",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "all_key_fields",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9012345679012346,
      "lo": 0.7347069741572956,
      "hi": 0.9678105945683666,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "line_items_f1",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9883720930232558,
      "lo": 0.9571428571428572,
      "hi": 1,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "extraction",
      "metric": "valid_json",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8754449702581328,
      "hi": 0.9999999999999998,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.93875,
      "lo": 0.8771796875,
      "hi": 0.98875,
      "n": 120,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.825,
      "lo": 0.7,
      "hi": 0.925,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.825,
      "lo": 0.7,
      "hi": 0.925,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.825,
      "lo": 0.7,
      "hi": 0.95,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8083333333333333,
      "lo": 0.6617124786829255,
      "hi": 0.9009208672454144,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8083333333333333,
      "lo": 0.6617124786829255,
      "hi": 0.9009208672454144,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6190476190476192,
      "lo": 0.3666263699227146,
      "hi": 0.8202043913302295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6190476190476192,
      "lo": 0.3666263699227146,
      "hi": 0.8202043913302295,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9275028768699655,
      "lo": 0.8503739930955121,
      "hi": 0.9879315304948215,
      "n": 112,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7878787878787878,
      "lo": 0.6363636363636364,
      "hi": 0.9090909090909091,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7878787878787878,
      "lo": 0.6363636363636364,
      "hi": 0.9090909090909091,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7878787878787878,
      "lo": 0.6363636363636364,
      "hi": 0.9393939393939394,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7676767676767676,
      "lo": 0.6005703146281053,
      "hi": 0.8789619089097739,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004219409282700422,
      "lo": 0.00031379493234937397,
      "hi": 0.05410491704516887,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7676767676767676,
      "lo": 0.6005703146281053,
      "hi": 0.8789619089097739,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9227777777777778,
      "lo": 0.8283333333333334,
      "hi": 0.9866666666666667,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8784722222222222,
      "lo": 0.6666666666666666,
      "hi": 1,
      "n": 26,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.037037037037037035,
      "lo": 0.004871638834261019,
      "hi": 0.23205356824597756,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.037037037037037035,
      "lo": 0.004871638834261019,
      "hi": 0.23205356824597756,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9337837837837838,
      "lo": 0.8672170608108108,
      "hi": 0.9891891891891892,
      "n": 117,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8108108108108109,
      "lo": 0.6756756756756757,
      "hi": 0.918918918918919,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8108108108108109,
      "lo": 0.6756756756756757,
      "hi": 0.918918918918919,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8108108108108109,
      "lo": 0.6756756756756757,
      "hi": 0.9459459459459459,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7927927927927929,
      "lo": 0.637936708579004,
      "hi": 0.8925699699847967,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7927927927927929,
      "lo": 0.637936708579004,
      "hi": 0.8925699699847967,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9444444444444443,
      "lo": 0.6806599518783086,
      "hi": 0.992678708872588,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9444444444444443,
      "lo": 0.6806599518783086,
      "hi": 0.992678708872588,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9232456140350878,
      "lo": 0.7785087719298246,
      "hi": 1,
      "n": 31,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8333333333333334,
      "lo": 0.5833333333333334,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8055555555555557,
      "lo": 0.5229574781220162,
      "hi": 0.9399628511442256,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.03508771929824561,
      "lo": 0.004614321029623267,
      "hi": 0.2219382908087743,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8055555555555557,
      "lo": 0.5229574781220162,
      "hi": 0.9399628511442256,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.03508771929824561,
      "lo": 0.004614321029623267,
      "hi": 0.2219382908087743,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9227777777777778,
      "lo": 0.8260972222222223,
      "hi": 0.9866666666666667,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.4,
      "hi": 0.9333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6444444444444445,
      "lo": 0.39685647827723214,
      "hi": 0.8331327903124117,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005555555555555555,
      "lo": 0.0004131987259021891,
      "hi": 0.07020118839238668,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.03333333333333333,
      "lo": 0.004382822599019884,
      "hi": 0.2126673249169449,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.03333333333333333,
      "lo": 0.004382822599019884,
      "hi": 0.2126673249169449,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.012345679012345678,
      "lo": 0.0009186195639853539,
      "hi": 0.14525233536944054,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7833333333333333,
      "lo": 0.5661276081063967,
      "hi": 0.909234802330339,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7833333333333333,
      "lo": 0.5661276081063967,
      "hi": 0.909234802330339,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8333333333333333,
      "lo": 0.6206812940840897,
      "hi": 0.938568600547364,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.8333333333333333,
      "lo": 0.6206812940840897,
      "hi": 0.938568600547364,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0125,
      "lo": 0.0022099755204522745,
      "hi": 0.06746270077668665,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.008333333333333333,
      "lo": 0.0010929336567140485,
      "hi": 0.06062822705322094,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 100,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.8388748419471806,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.84125,
      "lo": 0.615,
      "hi": 1,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7,
      "lo": 0.4,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.36779781217588203,
      "hi": 0.8730245878698215,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6666666666666666,
      "lo": 0.36779781217588203,
      "hi": 0.8730245878698215,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.91375,
      "lo": 0.795625,
      "hi": 0.989375,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6,
      "lo": 0.3,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6,
      "lo": 0.3,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6,
      "lo": 0.3,
      "hi": 1,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.5666666666666667,
      "lo": 0.28645461820349444,
      "hi": 0.8098743418147871,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9057692307692308,
      "lo": 0.806953125,
      "hi": 0.9798076923076923,
      "n": 106,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7307692307692307,
      "lo": 0.5384615384615384,
      "hi": 0.8846153846153846,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7307692307692307,
      "lo": 0.5384615384615384,
      "hi": 0.8846153846153846,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7307692307692307,
      "lo": 0.5384615384615384,
      "hi": 0.9230769230769231,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7307692307692307,
      "lo": 0.5391695962082381,
      "hi": 0.8629555167798162,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.7307692307692307,
      "lo": 0.5391695962082381,
      "hi": 0.8629555167798162,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 94,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 1,
      "hi": 1,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9523809523809523,
      "lo": 0.7162271925814832,
      "hi": 0.9937297001797036,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9523809523809523,
      "lo": 0.7162271925814832,
      "hi": 0.9937297001797036,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "decision_consistency",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.9666666666666667,
      "lo": 0.917419662440894,
      "hi": 0.9869624489016494,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "total_consistency",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "calibration_error",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.033444444444444194,
      "lo": 0.016888888888888596,
      "hi": 0.07222361111111085,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 402.04166666666663,
      "lo": 145.03854166666676,
      "hi": 764.4937499999996,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 331.91666666666663,
      "lo": 108.33333333333317,
      "hi": 662.5895833333325,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:gemini-3.8-flash",
      "value": 231.45416666666668,
      "lo": 13.333333333333343,
      "hi": 592.0166666666661,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:gemini-3.8-flash",
      "value": 160.69166666666666,
      "lo": 10.000000000000021,
      "hi": 450.42916666666656,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:gemini-3.8-flash",
      "value": 1160.2083333333333,
      "lo": 592.9166666666661,
      "hi": 1862.0833333333333,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:gemini-3.8-flash",
      "value": 1092.9166666666665,
      "lo": 500.000000000001,
      "hi": 1769.1666666666667,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_32usd",
      "target": "diy:gemini-3.8-flash",
      "value": 271.7083333333333,
      "lo": 42.66666666666669,
      "hi": 657,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_32usd",
      "target": "diy:gemini-3.8-flash",
      "value": 201.58333333333334,
      "lo": 34.666666666666615,
      "hi": 511.41666666666663,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_182usd",
      "target": "diy:gemini-3.8-flash",
      "value": 559.2083333333333,
      "lo": 236.72187500000027,
      "hi": 985.6666666666666,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_182usd",
      "target": "diy:gemini-3.8-flash",
      "value": 489.0833333333333,
      "lo": 197.16666666666637,
      "hi": 871.9999999999999,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_8usd",
      "target": "diy:gemini-3.8-flash",
      "value": 290.66666666666663,
      "lo": 124.65833333333325,
      "hi": 498.00000000000006,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_8usd",
      "target": "diy:gemini-3.8-flash",
      "value": 257.66666666666663,
      "lo": 100.0000000000002,
      "hi": 448.0166666666666,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_73usd",
      "target": "diy:gemini-3.8-flash",
      "value": 1095.0416666666667,
      "lo": 141.66666666666683,
      "hi": 2617.749999999998,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_73usd",
      "target": "diy:gemini-3.8-flash",
      "value": 793.9166666666667,
      "lo": 108.33333333333317,
      "hi": 2081.958333333332,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_0_1pct",
      "target": "diy:gemini-3.8-flash",
      "value": 100,
      "lo": 100,
      "hi": 100,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_5pct",
      "target": "diy:gemini-3.8-flash",
      "value": 5000,
      "lo": 5000,
      "hi": 5000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_32usd",
      "target": "diy:gemini-3.8-flash",
      "value": 320,
      "lo": 320,
      "hi": 320,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_182usd",
      "target": "diy:gemini-3.8-flash",
      "value": 1820,
      "lo": 1820,
      "hi": 1820,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_8usd",
      "target": "diy:gemini-3.8-flash",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_73usd",
      "target": "diy:gemini-3.8-flash",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "evidence",
      "metric": "edited_field_named",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.6166666666666666,
      "lo": 0.40180353972776706,
      "hi": 0.7939339233932416,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 6719,
      "lo": 6719,
      "hi": 6719,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 12793.349999999993,
      "lo": 12793.349999999993,
      "hi": 12793.349999999993,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 22307.53999999999,
      "lo": 22307.53999999999,
      "hi": 22307.53999999999,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004294021666666665,
      "lo": 0.004294021666666665,
      "hi": 0.004294021666666665,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004294021666666665,
      "lo": 0.004294021666666665,
      "hi": 0.004294021666666665,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "all",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 6544,
      "lo": 6544,
      "hi": 6544,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 11545.899999999998,
      "lo": 11545.899999999998,
      "hi": 11545.899999999998,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 14641.350000000019,
      "lo": 14641.350000000019,
      "hi": 14641.350000000019,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004114682539682538,
      "lo": 0.004114682539682538,
      "hi": 0.004114682539682538,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004114682539682538,
      "lo": 0.004114682539682538,
      "hi": 0.004114682539682538,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=consistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 9947,
      "lo": 9947,
      "hi": 9947,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 21572.449999999986,
      "lo": 21572.449999999986,
      "hi": 21572.449999999986,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 24224.68999999999,
      "lo": 24224.68999999999,
      "hi": 24224.68999999999,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.006269767857142859,
      "lo": 0.006269767857142859,
      "hi": 0.006269767857142859,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0.006269767857142859,
      "lo": 0.006269767857142859,
      "hi": 0.006269767857142859,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 6156,
      "lo": 6156,
      "hi": 6156,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 13547.099999999997,
      "lo": 13547.099999999997,
      "hi": 13547.099999999997,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 30896.08000000002,
      "lo": 30896.08000000002,
      "hi": 30896.08000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0037876500000000005,
      "lo": 0.0037876500000000005,
      "hi": 0.0037876500000000005,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0037876500000000005,
      "lo": 0.0037876500000000005,
      "hi": 0.0037876500000000005,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=unknown",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 7598,
      "lo": 7598,
      "hi": 7598,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 16502.44999999998,
      "lo": 16502.44999999998,
      "hi": 16502.44999999998,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 24224.68999999999,
      "lo": 24224.68999999999,
      "hi": 24224.68999999999,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004773017857142857,
      "lo": 0.004773017857142857,
      "hi": 0.004773017857142857,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004773017857142857,
      "lo": 0.004773017857142857,
      "hi": 0.004773017857142857,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 10323.5,
      "lo": 10323.5,
      "hi": 10323.5,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 20283.349999999984,
      "lo": 20283.349999999984,
      "hi": 20283.349999999984,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 27062.269999999982,
      "lo": 27062.269999999982,
      "hi": 27062.269999999982,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005811833333333334,
      "lo": 0.005811833333333334,
      "hi": 0.005811833333333334,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005811833333333334,
      "lo": 0.005811833333333334,
      "hi": 0.005811833333333334,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 6566.5,
      "lo": 6566.5,
      "hi": 6566.5,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 12470.049999999996,
      "lo": 12470.049999999996,
      "hi": 12470.049999999996,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 22587.05000000003,
      "lo": 22587.05000000003,
      "hi": 22587.05000000003,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004194297794117645,
      "lo": 0.004194297794117645,
      "hi": 0.004194297794117645,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004194297794117645,
      "lo": 0.004194297794117645,
      "hi": 0.004194297794117645,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=photo",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 8382.5,
      "lo": 8382.5,
      "hi": 8382.5,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 17063.549999999996,
      "lo": 17063.549999999996,
      "hi": 17063.549999999996,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 20713.899999999998,
      "lo": 20713.899999999998,
      "hi": 20713.899999999998,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0057410312500000005,
      "lo": 0.0057410312500000005,
      "hi": 0.0057410312500000005,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0057410312500000005,
      "lo": 0.0057410312500000005,
      "hi": 0.0057410312500000005,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=scan",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 7554,
      "lo": 7554,
      "hi": 7554,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 10726.299999999997,
      "lo": 10726.299999999997,
      "hi": 10726.299999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 11925.259999999998,
      "lo": 11925.259999999998,
      "hi": 11925.259999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00460375,
      "lo": 0.00460375,
      "hi": 0.00460375,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00460375,
      "lo": 0.00460375,
      "hi": 0.00460375,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 9146.5,
      "lo": 9146.5,
      "hi": 9146.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 12773.799999999996,
      "lo": 12773.799999999996,
      "hi": 12773.799999999996,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 19140.699999999986,
      "lo": 19140.699999999986,
      "hi": 19140.699999999986,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0056297375,
      "lo": 0.0056297375,
      "hi": 0.0056297375,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0056297375,
      "lo": 0.0056297375,
      "hi": 0.0056297375,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_full",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 9294.5,
      "lo": 9294.5,
      "hi": 9294.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 17578.399999999994,
      "lo": 17578.399999999994,
      "hi": 17578.399999999994,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 25865.99000000001,
      "lo": 25865.99000000001,
      "hi": 25865.99000000001,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0053229750000000015,
      "lo": 0.0053229750000000015,
      "hi": 0.0053229750000000015,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0053229750000000015,
      "lo": 0.0053229750000000015,
      "hi": 0.0053229750000000015,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_inpaint",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 6649,
      "lo": 6649,
      "hi": 6649,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 10752.35,
      "lo": 10752.35,
      "hi": 10752.35,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 12579.309999999998,
      "lo": 12579.309999999998,
      "hi": 12579.309999999998,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004236741666666667,
      "lo": 0.004236741666666667,
      "hi": 0.004236741666666667,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004236741666666667,
      "lo": 0.004236741666666667,
      "hi": 0.004236741666666667,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=duplicate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 6236,
      "lo": 6236,
      "hi": 6236,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 11100.999999999998,
      "lo": 11100.999999999998,
      "hi": 11100.999999999998,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 14383.3,
      "lo": 14383.3,
      "hi": 14383.3,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038457644927536224,
      "lo": 0.0038457644927536224,
      "hi": 0.0038457644927536224,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038457644927536224,
      "lo": 0.0038457644927536224,
      "hi": 0.0038457644927536224,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=genuine",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 5970,
      "lo": 5970,
      "hi": 5970,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 10580.399999999994,
      "lo": 10580.399999999994,
      "hi": 10580.399999999994,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 14908.039999999999,
      "lo": 14908.039999999999,
      "hi": 14908.039999999999,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0033959318181818188,
      "lo": 0.0033959318181818188,
      "hi": 0.0033959318181818188,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0033959318181818188,
      "lo": 0.0033959318181818188,
      "hi": 0.0033959318181818188,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=hard_negative",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 7497,
      "lo": 7497,
      "hi": 7497,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 20118.99999999998,
      "lo": 20118.99999999998,
      "hi": 20118.99999999998,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 24499.61,
      "lo": 24499.61,
      "hi": 24499.61,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00484635,
      "lo": 0.00484635,
      "hi": 0.00484635,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00484635,
      "lo": 0.00484635,
      "hi": 0.00484635,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=manual_edit",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 9018,
      "lo": 9018,
      "hi": 9018,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 24124.249999999978,
      "lo": 24124.249999999978,
      "hi": 24124.249999999978,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 35028.049999999974,
      "lo": 35028.049999999974,
      "hi": 35028.049999999974,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005680875000000001,
      "lo": 0.005680875000000001,
      "hi": 0.005680875000000001,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005680875000000001,
      "lo": 0.005680875000000001,
      "hi": 0.005680875000000001,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=EUR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 6510,
      "lo": 6510,
      "hi": 6510,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 11101.799999999996,
      "lo": 11101.799999999996,
      "hi": 11101.799999999996,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 16035.879999999972,
      "lo": 16035.879999999972,
      "hi": 16035.879999999972,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004010708791208793,
      "lo": 0.004010708791208793,
      "hi": 0.004010708791208793,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004010708791208793,
      "lo": 0.004010708791208793,
      "hi": 0.004010708791208793,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=IDR",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 6278,
      "lo": 6278,
      "hi": 6278,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 13703.5,
      "lo": 13703.5,
      "hi": 13703.5,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 26862.89000000002,
      "lo": 26862.89000000002,
      "hi": 26862.89000000002,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00432938125,
      "lo": 0.00432938125,
      "hi": 0.00432938125,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00432938125,
      "lo": 0.00432938125,
      "hi": 0.00432938125,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=USD",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 6675,
      "lo": 6675,
      "hi": 6675,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 12864,
      "lo": 12864,
      "hi": 12864,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 22411.400000000027,
      "lo": 22411.400000000027,
      "hi": 22411.400000000027,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004276629251700679,
      "lo": 0.004276629251700679,
      "hi": 0.004276629251700679,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004276629251700679,
      "lo": 0.004276629251700679,
      "hi": 0.004276629251700679,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "docType=receipt",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 6171,
      "lo": 6171,
      "hi": 6171,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 8039.099999999998,
      "lo": 8039.099999999998,
      "hi": 8039.099999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 9245.419999999996,
      "lo": 9245.419999999996,
      "hi": 9245.419999999996,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.003508833333333334,
      "lo": 0.003508833333333334,
      "hi": 0.003508833333333334,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.003508833333333334,
      "lo": 0.003508833333333334,
      "hi": 0.003508833333333334,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=items+total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 9947,
      "lo": 9947,
      "hi": 9947,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 19809.25,
      "lo": 19809.25,
      "hi": 19809.25,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 24362.149999999998,
      "lo": 24362.149999999998,
      "hi": 24362.149999999998,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.006104541666666668,
      "lo": 0.006104541666666668,
      "hi": 0.006104541666666668,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0.006104541666666668,
      "lo": 0.006104541666666668,
      "hi": 0.006104541666666668,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=total",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 6500,
      "lo": 6500,
      "hi": 6500,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 13689.2,
      "lo": 13689.2,
      "hi": 13689.2,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 25367.539999999964,
      "lo": 25367.539999999964,
      "hi": 25367.539999999964,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004454044444444443,
      "lo": 0.004454044444444443,
      "hi": 0.004454044444444443,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004454044444444443,
      "lo": 0.004454044444444443,
      "hi": 0.004454044444444443,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=en",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 6510,
      "lo": 6510,
      "hi": 6510,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 11101.799999999996,
      "lo": 11101.799999999996,
      "hi": 11101.799999999996,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 16035.879999999972,
      "lo": 16035.879999999972,
      "hi": 16035.879999999972,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004010708791208793,
      "lo": 0.004010708791208793,
      "hi": 0.004010708791208793,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004010708791208793,
      "lo": 0.004010708791208793,
      "hi": 0.004010708791208793,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=id",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 8516.5,
      "lo": 8516.5,
      "hi": 8516.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 12171.699999999999,
      "lo": 12171.699999999999,
      "hi": 12171.699999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 12327.16,
      "lo": 12327.16,
      "hi": 12327.16,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005138675,
      "lo": 0.005138675,
      "hi": 0.005138675,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005138675,
      "lo": 0.005138675,
      "hi": 0.005138675,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 9673,
      "lo": 9673,
      "hi": 9673,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 15861.29999999999,
      "lo": 15861.29999999999,
      "hi": 15861.29999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 20451.700000000004,
      "lo": 20451.700000000004,
      "hi": 20451.700000000004,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0061208,
      "lo": 0.0061208,
      "hi": 0.0061208,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0061208,
      "lo": 0.0061208,
      "hi": 0.0061208,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=free",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 9557,
      "lo": 9557,
      "hi": 9557,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 21776.699999999997,
      "lo": 21776.699999999997,
      "hi": 21776.699999999997,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 26637.129999999986,
      "lo": 26637.129999999986,
      "hi": 26637.129999999986,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005572362500000001,
      "lo": 0.005572362500000001,
      "hi": 0.005572362500000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005572362500000001,
      "lo": 0.005572362500000001,
      "hi": 0.005572362500000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=intact",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 6265.5,
      "lo": 6265.5,
      "hi": 6265.5,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 11009.35,
      "lo": 11009.35,
      "hi": 11009.35,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 15142.949999999979,
      "lo": 15142.949999999979,
      "hi": 15142.949999999979,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.003907411363636362,
      "lo": 0.003907411363636362,
      "hi": 0.003907411363636362,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.003907411363636362,
      "lo": 0.003907411363636362,
      "hi": 0.003907411363636362,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 8262.5,
      "lo": 8262.5,
      "hi": 8262.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 12421.649999999996,
      "lo": 12421.649999999996,
      "hi": 12421.649999999996,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 17202.07999999999,
      "lo": 17202.07999999999,
      "hi": 17202.07999999999,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005142037500000002,
      "lo": 0.005142037500000002,
      "hi": 0.005142037500000002,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005142037500000002,
      "lo": 0.005142037500000002,
      "hi": 0.005142037500000002,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=stripped",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 6320,
      "lo": 6320,
      "hi": 6320,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 10531.799999999985,
      "lo": 10531.799999999985,
      "hi": 10531.799999999985,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 13924.289999999994,
      "lo": 13924.289999999994,
      "hi": 13924.289999999994,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038432631578947394,
      "lo": 0.0038432631578947394,
      "hi": 0.0038432631578947394,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038432631578947394,
      "lo": 0.0038432631578947394,
      "hi": 0.0038432631578947394,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=cord",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 9822,
      "lo": 9822,
      "hi": 9822,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 16906.399999999998,
      "lo": 16906.399999999998,
      "hi": 16906.399999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 18411.68,
      "lo": 18411.68,
      "hi": 18411.68,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0057788,
      "lo": 0.0057788,
      "hi": 0.0057788,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0057788,
      "lo": 0.0057788,
      "hi": 0.0057788,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 8635,
      "lo": 8635,
      "hi": 8635,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 18372.499999999985,
      "lo": 18372.499999999985,
      "hi": 18372.499999999985,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 26680.09999999999,
      "lo": 26680.09999999999,
      "hi": 26680.09999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004867150000000001,
      "lo": 0.004867150000000001,
      "hi": 0.004867150000000001,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004867150000000001,
      "lo": 0.004867150000000001,
      "hi": 0.004867150000000001,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 7497,
      "lo": 7497,
      "hi": 7497,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 20118.99999999998,
      "lo": 20118.99999999998,
      "hi": 20118.99999999998,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 24499.61,
      "lo": 24499.61,
      "hi": 24499.61,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00484635,
      "lo": 0.00484635,
      "hi": 0.00484635,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00484635,
      "lo": 0.00484635,
      "hi": 0.00484635,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 6145,
      "lo": 6145,
      "hi": 6145,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 11622.549999999997,
      "lo": 11622.549999999997,
      "hi": 11622.549999999997,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 30923.65999999992,
      "lo": 30923.65999999992,
      "hi": 30923.65999999992,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004050801470588235,
      "lo": 0.004050801470588235,
      "hi": 0.004050801470588235,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004050801470588235,
      "lo": 0.004050801470588235,
      "hi": 0.004050801470588235,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=expressexpense",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 9874.5,
      "lo": 9874.5,
      "hi": 9874.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 12452.35,
      "lo": 12452.35,
      "hi": 12452.35,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 12656.25,
      "lo": 12656.25,
      "hi": 12656.25,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005726599999999999,
      "lo": 0.005726599999999999,
      "hi": 0.005726599999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0.005726599999999999,
      "lo": 0.005726599999999999,
      "hi": 0.005726599999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 8674,
      "lo": 8674,
      "hi": 8674,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 15861.29999999999,
      "lo": 15861.29999999999,
      "hi": 15861.29999999999,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 20451.700000000004,
      "lo": 20451.700000000004,
      "hi": 20451.700000000004,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0055328750000000005,
      "lo": 0.0055328750000000005,
      "hi": 0.0055328750000000005,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0055328750000000005,
      "lo": 0.0055328750000000005,
      "hi": 0.0055328750000000005,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 6228,
      "lo": 6228,
      "hi": 6228,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 9199.799999999997,
      "lo": 9199.799999999997,
      "hi": 9199.799999999997,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 12732.000000000007,
      "lo": 12732.000000000007,
      "hi": 12732.000000000007,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038509102564102567,
      "lo": 0.0038509102564102567,
      "hi": 0.0038509102564102567,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038509102564102567,
      "lo": 0.0038509102564102567,
      "hi": 0.0038509102564102567,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=exclusive",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 6628,
      "lo": 6628,
      "hi": 6628,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 10785.699999999999,
      "lo": 10785.699999999999,
      "hi": 10785.699999999999,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 13479.47999999999,
      "lo": 13479.47999999999,
      "hi": 13479.47999999999,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038343409090909088,
      "lo": 0.0038343409090909088,
      "hi": 0.0038343409090909088,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0038343409090909088,
      "lo": 0.0038343409090909088,
      "hi": 0.0038343409090909088,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=none",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 5848,
      "lo": 5848,
      "hi": 5848,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 9631.699999999999,
      "lo": 9631.699999999999,
      "hi": 9631.699999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 10549.539999999999,
      "lo": 10549.539999999999,
      "hi": 10549.539999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0039997,
      "lo": 0.0039997,
      "hi": 0.0039997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0039997,
      "lo": 0.0039997,
      "hi": 0.0039997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=crop",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 6353,
      "lo": 6353,
      "hi": 6353,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 9126.5,
      "lo": 9126.5,
      "hi": 9126.5,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 9577.3,
      "lo": 9577.3,
      "hi": 9577.3,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00387685,
      "lo": 0.00387685,
      "hi": 0.00387685,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00387685,
      "lo": 0.00387685,
      "hi": 0.00387685,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=jpeg",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 6276,
      "lo": 6276,
      "hi": 6276,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 10930.199999999999,
      "lo": 10930.199999999999,
      "hi": 10930.199999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 11279.64,
      "lo": 11279.64,
      "hi": 11279.64,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004219049999999999,
      "lo": 0.004219049999999999,
      "hi": 0.004219049999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004219049999999999,
      "lo": 0.004219049999999999,
      "hi": 0.004219049999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rephoto",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 6090,
      "lo": 6090,
      "hi": 6090,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 11986.599999999995,
      "lo": 11986.599999999995,
      "hi": 11986.599999999995,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 14754.119999999997,
      "lo": 14754.119999999997,
      "hi": 14754.119999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0045724500000000005,
      "lo": 0.0045724500000000005,
      "hi": 0.0045724500000000005,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 0.0045724500000000005,
      "lo": 0.0045724500000000005,
      "hi": 0.0045724500000000005,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=resize",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 6255,
      "lo": 6255,
      "hi": 6255,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 9567.899999999998,
      "lo": 9567.899999999998,
      "hi": 9567.899999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 10521.579999999998,
      "lo": 10521.579999999998,
      "hi": 10521.579999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004148649999999999,
      "lo": 0.004148649999999999,
      "hi": 0.004148649999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 0.004148649999999999,
      "lo": 0.004148649999999999,
      "hi": 0.004148649999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rotate",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 7554,
      "lo": 7554,
      "hi": 7554,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 10726.299999999997,
      "lo": 10726.299999999997,
      "hi": 10726.299999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 11925.259999999998,
      "lo": 11925.259999999998,
      "hi": 11925.259999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00460375,
      "lo": 0.00460375,
      "hi": 0.00460375,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0.00460375,
      "lo": 0.00460375,
      "hi": 0.00460375,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=screenshot",
      "target": "diy:gemini-3.8-flash",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.91078125,
      "lo": 0.83953125,
      "hi": 0.96625,
      "n": 120,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.65,
      "lo": 0.525,
      "hi": 0.85,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.75,
      "lo": 0.625,
      "hi": 0.9,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.8,
      "lo": 0.65,
      "hi": 0.925,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.65,
      "lo": 0.49505880837257693,
      "hi": 0.7786547112682372,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.55,
      "lo": 0.39829091798932054,
      "hi": 0.6929469218909508,
      "n": 40,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.35714285714285715,
      "lo": 0.1634473175355414,
      "hi": 0.612355768960926,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.14285714285714285,
      "lo": 0.040093921271464966,
      "hi": 0.39941379496970353,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.8908707326428845,
      "lo": 0.8112677406981205,
      "hi": 0.9595320291522823,
      "n": 112,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.5757575757575758,
      "lo": 0.42424242424242425,
      "hi": 0.8181818181818182,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.696969696969697,
      "lo": 0.5151515151515151,
      "hi": 0.8484848484848485,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.7575757575757576,
      "lo": 0.5757575757575758,
      "hi": 0.8787878787878788,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.5757575757575758,
      "lo": 0.4080726333731245,
      "hi": 0.7276440336658979,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.45454545454545453,
      "lo": 0.29842938513876116,
      "hi": 0.6201406146378254,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.04637121165392266,
      "n": 79,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.845,
      "lo": 0.7199722222222222,
      "hi": 0.9505694444444444,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.4,
      "lo": 0.2,
      "hi": 0.7333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.5333333333333333,
      "lo": 0.26666666666666666,
      "hi": 0.8,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3333333333333333,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.4,
      "lo": 0.1982449614134158,
      "hi": 0.6425316987937539,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.2,
      "lo": 0.07047549346981563,
      "hi": 0.45185448715169363,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0.9444444444444444,
      "lo": 0.8125,
      "hi": 1,
      "n": 26,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.9035472972972973,
      "lo": 0.8315836148648649,
      "hi": 0.9628420608108108,
      "n": 117,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.6216216216216216,
      "lo": 0.4864864864864865,
      "hi": 0.8378378378378378,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.7297297297297297,
      "lo": 0.5945945945945946,
      "hi": 0.8648648648648649,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.7837837837837838,
      "lo": 0.6216216216216216,
      "hi": 0.918918918918919,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.6216216216216216,
      "lo": 0.46100430439217654,
      "hi": 0.7593600082420173,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.5135135135135135,
      "lo": 0.358948966656188,
      "hi": 0.6655359569698336,
      "n": 37,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.2537815976337061,
      "hi": 0.7462184023662939,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0.25,
      "lo": 0.08894166839405471,
      "hi": 0.5323053349335657,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.9605263157894737,
      "lo": 0.868421052631579,
      "hi": 1,
      "n": 31,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.9166666666666666,
      "lo": 0.75,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.9166666666666666,
      "lo": 0.75,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.9166666666666666,
      "lo": 0.75,
      "hi": 1,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.75,
      "lo": 0.46769466506643426,
      "hi": 0.9110583316059453,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.6666666666666666,
      "lo": 0.39062208887279953,
      "hi": 0.8618799089087867,
      "n": 12,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.16817922405261626,
      "n": 19,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.845,
      "lo": 0.7055416666666667,
      "hi": 0.9494444444444444,
      "n": 75,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.4,
      "lo": 0.2,
      "hi": 0.7333333333333333,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.5333333333333333,
      "lo": 0.26666666666666666,
      "hi": 0.8,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3333333333333333,
      "hi": 0.8666666666666667,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.4,
      "lo": 0.1982449614134158,
      "hi": 0.6425316987937539,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.2,
      "lo": 0.07047549346981563,
      "hi": 0.45185448715169363,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.5958499732047615,
      "hi": 0.9821237869049271,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.236593090512564,
      "hi": 0.7634069094874361,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0.2,
      "lo": 0.05668215145437519,
      "hi": 0.5098375284633583,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.16112515805281938,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 1,
      "lo": 0.7224672001371107,
      "hi": 0.9999999999999999,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0.8,
      "lo": 0.49016247153664183,
      "hi": 0.9433178485456247,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11696979849974075,
      "n": 29,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.12455502974186705,
      "n": 27,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3865815007622531,
      "hi": 0.781193467627183,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.2992980081982123,
      "hi": 0.7007019918017877,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0.7,
      "lo": 0.4810271816464765,
      "hi": 0.8545227551323956,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3865815007622531,
      "hi": 0.781193467627183,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_5",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "false_flag_rate_at_risk_0_7",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.0458181295355271,
      "n": 80,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.9825,
      "lo": 0.9428125,
      "hi": 1,
      "n": 100,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.75,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.95,
      "lo": 0.85,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.95,
      "lo": 0.85,
      "hi": 1,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.6989663547715128,
      "hi": 0.9721335187862319,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.9,
      "lo": 0.6989663547715128,
      "hi": 0.9721335187862319,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.77375,
      "lo": 0.557484375,
      "hi": 0.9475,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.3,
      "lo": 0.1,
      "hi": 0.7,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.2,
      "hi": 0.8,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.3,
      "lo": 0.10779126740630099,
      "hi": 0.6032218525388546,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.2,
      "lo": 0.05668215145437519,
      "hi": 0.5098375284633583,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.904375,
      "lo": 0.8012187500000001,
      "hi": 0.979375,
      "n": 90,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.2,
      "hi": 0.8,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.6,
      "lo": 0.3,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.7,
      "lo": 0.4,
      "hi": 0.9,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.5,
      "lo": 0.236593090512564,
      "hi": 0.7634069094874361,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.2,
      "lo": 0.05668215145437519,
      "hi": 0.5098375284633583,
      "n": 10,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.9132211538461539,
      "lo": 0.8204026442307694,
      "hi": 0.9841466346153845,
      "n": 106,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.6923076923076923,
      "lo": 0.5375000000000016,
      "hi": 0.9230769230769231,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.8076923076923077,
      "lo": 0.6538461538461539,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.8461538461538461,
      "lo": 0.6913461538461555,
      "hi": 0.9615384615384616,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.6923076923076923,
      "lo": 0.5001174167780164,
      "hi": 0.8349868440453624,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.6538461538461539,
      "lo": 0.46220571188810816,
      "hi": 0.8058776967705947,
      "n": 26,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "auc",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.90625,
      "lo": 0.8080357142857143,
      "hi": 0.9794642857142857,
      "n": 94,
      "unit": "auc",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_1",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.5714285714285714,
      "lo": 0.35714285714285715,
      "hi": 0.8571428571428571,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_5",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.6428571428571429,
      "lo": 0.42857142857142855,
      "hi": 0.9285714285714286,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_fpr_10",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.7142857142857143,
      "lo": 0.49821428571428883,
      "hi": 0.9285714285714286,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_5",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.5714285714285714,
      "lo": 0.3259064457965074,
      "hi": 0.7861920109552588,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "authenticity",
      "metric": "detection_at_risk_0_7",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.35714285714285715,
      "lo": 0.1634473175355414,
      "hi": 0.612355768960926,
      "n": 14,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "decision_consistency",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "total_consistency",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 1,
      "lo": 0.9689808335812966,
      "hi": 1,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "decisions",
      "metric": "calibration_error",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.06450000000000017,
      "lo": 0.044495833333333366,
      "hi": 0.12175625000000015,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 350,
      "lo": 199.99999999999994,
      "hi": 500,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 449.99999999999994,
      "lo": 300.00000000000006,
      "hi": 600,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_0_1pct",
      "target": "decision:gpt-6-luna",
      "value": 34.99999999999999,
      "lo": 19.999999999999996,
      "hi": 50,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_0_1pct",
      "target": "decision:gpt-6-luna",
      "value": 45,
      "lo": 30.000000000000004,
      "hi": 60,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fakes_5pct",
      "target": "decision:gpt-6-luna",
      "value": 1749.9999999999998,
      "lo": 999.9999999999999,
      "hi": 2500,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fakes_5pct",
      "target": "decision:gpt-6-luna",
      "value": 2250,
      "lo": 1500.0000000000002,
      "hi": 3000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_32usd",
      "target": "decision:gpt-6-luna",
      "value": 111.99999999999999,
      "lo": 63.999999999999986,
      "hi": 160,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_32usd",
      "target": "decision:gpt-6-luna",
      "value": 144,
      "lo": 96.00000000000001,
      "hi": 192,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=fake_182usd",
      "target": "decision:gpt-6-luna",
      "value": 636.9999999999999,
      "lo": 363.99999999999994,
      "hi": 910,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=fake_182usd",
      "target": "decision:gpt-6-luna",
      "value": 819,
      "lo": 546,
      "hi": 1092,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_8usd",
      "target": "decision:gpt-6-luna",
      "value": 350,
      "lo": 199.99999999999994,
      "hi": 500,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_8usd",
      "target": "decision:gpt-6-luna",
      "value": 449.99999999999994,
      "lo": 300.00000000000006,
      "hi": 600,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_5",
      "stratum": "assumption=review_73usd",
      "target": "decision:gpt-6-luna",
      "value": 350,
      "lo": 199.99999999999994,
      "hi": 500,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "expected_cost_per_1000_at_risk_0_7",
      "stratum": "assumption=review_73usd",
      "target": "decision:gpt-6-luna",
      "value": 449.99999999999994,
      "lo": 300.00000000000006,
      "hi": 600,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_0_1pct",
      "target": "decision:gpt-6-luna",
      "value": 100,
      "lo": 100,
      "hi": 100,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fakes_5pct",
      "target": "decision:gpt-6-luna",
      "value": 5000,
      "lo": 5000,
      "hi": 5000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_32usd",
      "target": "decision:gpt-6-luna",
      "value": 320,
      "lo": 320,
      "hi": 320,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=fake_182usd",
      "target": "decision:gpt-6-luna",
      "value": 1820,
      "lo": 1820,
      "hi": 1820,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_8usd",
      "target": "decision:gpt-6-luna",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "decisions",
      "metric": "approve_all_cost_per_1000",
      "stratum": "assumption=review_73usd",
      "target": "decision:gpt-6-luna",
      "value": 1000,
      "lo": 1000,
      "hi": 1000,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "evidence",
      "metric": "edited_field_named",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.4,
      "lo": 0.21880653237281705,
      "hi": 0.6134184992377469,
      "n": 20,
      "unit": "rate",
      "higherIsBetter": true
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 364.5,
      "lo": 364.5,
      "hi": 364.5,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 737.1999999999998,
      "lo": 737.1999999999998,
      "hi": 737.1999999999998,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 1315.2899999999984,
      "lo": 1315.2899999999984,
      "hi": 1315.2899999999984,
      "n": 450,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.0001585466666666668,
      "lo": 0.0001585466666666668,
      "hi": 0.0001585466666666668,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0.0001585466666666668,
      "lo": 0.0001585466666666668,
      "hi": 0.0001585466666666668,
      "n": 450,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "all",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 8.673617379884035e-19,
      "hi": 0.008464318862970662,
      "n": 450,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 361,
      "lo": 361,
      "hi": 361,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 741.2499999999997,
      "lo": 741.2499999999997,
      "hi": 741.2499999999997,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 1417.9500000000012,
      "lo": 1417.9500000000012,
      "hi": 1417.9500000000012,
      "n": 378,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.00015688095238095256,
      "lo": 0.00015688095238095256,
      "hi": 0.00015688095238095256,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0.00015688095238095256,
      "lo": 0.00015688095238095256,
      "hi": 0.00015688095238095256,
      "n": 378,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=consistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01006035026306037,
      "n": 378,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 388.5,
      "lo": 388.5,
      "hi": 388.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 716.2999999999993,
      "lo": 716.2999999999993,
      "hi": 716.2999999999993,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 755.9499999999999,
      "lo": 755.9499999999999,
      "hi": 755.9499999999999,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.0001840714285714286,
      "lo": 0.0001840714285714286,
      "hi": 0.0001840714285714286,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0.0001840714285714286,
      "lo": 0.0001840714285714286,
      "hi": 0.0001840714285714286,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=inconsistent",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 347,
      "lo": 347,
      "hi": 347,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 666.5499999999998,
      "lo": 666.5499999999998,
      "hi": 666.5499999999998,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 992.9100000000003,
      "lo": 992.9100000000003,
      "hi": 992.9100000000003,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 0.00014380000000000003,
      "lo": 0.00014380000000000003,
      "hi": 0.00014380000000000003,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 0.00014380000000000003,
      "lo": 0.00014380000000000003,
      "hi": 0.00014380000000000003,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "arithmetic=unknown",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 362.5,
      "lo": 362.5,
      "hi": 362.5,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 514.2499999999998,
      "lo": 514.2499999999998,
      "hi": 514.2499999999998,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 653.6599999999994,
      "lo": 653.6599999999994,
      "hi": 653.6599999999994,
      "n": 42,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.00017085714285714295,
      "lo": 0.00017085714285714295,
      "hi": 0.00017085714285714295,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.00017085714285714295,
      "lo": 0.00017085714285714295,
      "hi": 0.00017085714285714295,
      "n": 42,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.08379879086570391,
      "n": 42,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 378.5,
      "lo": 378.5,
      "hi": 378.5,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 741.85,
      "lo": 741.85,
      "hi": 741.85,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 754.77,
      "lo": 754.77,
      "hi": 754.77,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0.00018533333333333333,
      "lo": 0.00018533333333333333,
      "hi": 0.00018533333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0.00018533333333333333,
      "lo": 0.00018533333333333333,
      "hi": 0.00018533333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "base=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 360.5,
      "lo": 360.5,
      "hi": 360.5,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 731.4999999999998,
      "lo": 731.4999999999998,
      "hi": 731.4999999999998,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 1222.3200000000002,
      "lo": 1222.3200000000002,
      "hi": 1222.3200000000002,
      "n": 408,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.00015812500000000004,
      "lo": 0.00015812500000000004,
      "hi": 0.00015812500000000004,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0.00015812500000000004,
      "lo": 0.00015812500000000004,
      "hi": 0.00015812500000000004,
      "n": 408,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=photo",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.009327518486589774,
      "n": 408,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 351,
      "lo": 351,
      "hi": 351,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 419.85,
      "lo": 419.85,
      "hi": 419.85,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 676.4099999999999,
      "lo": 676.4099999999999,
      "hi": 676.4099999999999,
      "n": 24,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 0.00016174999999999997,
      "lo": 0.00016174999999999997,
      "hi": 0.00016174999999999997,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 0.00016174999999999997,
      "lo": 0.00016174999999999997,
      "hi": 0.00016174999999999997,
      "n": 24,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=scan",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 1.3877787807814457e-17,
      "hi": 0.13797620467498017,
      "n": 24,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 440,
      "lo": 440,
      "hi": 440,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 1008.7999999999993,
      "lo": 1008.7999999999993,
      "hi": 1008.7999999999993,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 1376.1599999999996,
      "lo": 1376.1599999999996,
      "hi": 1376.1599999999996,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0.00015500000000000003,
      "lo": 0.00015500000000000003,
      "hi": 0.00015500000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0.00015500000000000003,
      "lo": 0.00015500000000000003,
      "hi": 0.00015500000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "capture=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 406,
      "lo": 406,
      "hi": 406,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 949.6499999999994,
      "lo": 949.6499999999994,
      "hi": 949.6499999999994,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 1368.9199999999987,
      "lo": 1368.9199999999987,
      "hi": 1368.9199999999987,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.0001849000000000001,
      "lo": 0.0001849000000000001,
      "hi": 0.0001849000000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0.0001849000000000001,
      "lo": 0.0001849000000000001,
      "hi": 0.0001849000000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_full",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 376.5,
      "lo": 376.5,
      "hi": 376.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 647.6499999999994,
      "lo": 647.6499999999994,
      "hi": 647.6499999999994,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 752.49,
      "lo": 752.49,
      "hi": 752.49,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.00018549999999999993,
      "lo": 0.00018549999999999993,
      "hi": 0.00018549999999999993,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0.00018549999999999993,
      "lo": 0.00018549999999999993,
      "hi": 0.00018549999999999993,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=ai_inpaint",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 335,
      "lo": 335,
      "hi": 335,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 538.2,
      "lo": 538.2,
      "hi": 538.2,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 1410.1499999999999,
      "lo": 1410.1499999999999,
      "hi": 1410.1499999999999,
      "n": 90,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 0.00014496666666666654,
      "lo": 0.00014496666666666654,
      "hi": 0.00014496666666666654,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 0.00014496666666666654,
      "lo": 0.00014496666666666654,
      "hi": 0.00014496666666666654,
      "n": 90,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=duplicate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.04093562556432678,
      "n": 90,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 356,
      "lo": 356,
      "hi": 356,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 749.6999999999998,
      "lo": 749.6999999999998,
      "hi": 749.6999999999998,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 1217.4599999999998,
      "lo": 1217.4599999999998,
      "hi": 1217.4599999999998,
      "n": 207,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 0.0001531739130434782,
      "lo": 0.0001531739130434782,
      "hi": 0.0001531739130434782,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 0.0001531739130434782,
      "lo": 0.0001531739130434782,
      "hi": 0.0001531739130434782,
      "n": 207,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=genuine",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.018219655859813683,
      "n": 207,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 335,
      "lo": 335,
      "hi": 335,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 516.4,
      "lo": 516.4,
      "hi": 516.4,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 556.36,
      "lo": 556.36,
      "hi": 556.36,
      "n": 33,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 0.00015109090909090912,
      "lo": 0.00015109090909090912,
      "hi": 0.00015109090909090912,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 0.00015109090909090912,
      "lo": 0.00015109090909090912,
      "hi": 0.00015109090909090912,
      "n": 33,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=hard_negative",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.10426999754245206,
      "n": 33,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 359.5,
      "lo": 359.5,
      "hi": 359.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 510.94999999999976,
      "lo": 510.94999999999976,
      "hi": 510.94999999999976,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 674.5400000000002,
      "lo": 674.5400000000002,
      "hi": 674.5400000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.00016490000000000002,
      "lo": 0.00016490000000000002,
      "hi": 0.00016490000000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0.00016490000000000002,
      "lo": 0.00016490000000000002,
      "hi": 0.00016490000000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "class=manual_edit",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 389,
      "lo": 389,
      "hi": 389,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 518.7499999999998,
      "lo": 518.7499999999998,
      "hi": 518.7499999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 617.3499999999998,
      "lo": 617.3499999999998,
      "hi": 617.3499999999998,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 0.00017199999999999998,
      "lo": 0.00017199999999999998,
      "hi": 0.00017199999999999998,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 0.00017199999999999998,
      "lo": 0.00017199999999999998,
      "hi": 0.00017199999999999998,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=EUR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 365,
      "lo": 365,
      "hi": 365,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 716.5999999999997,
      "lo": 716.5999999999997,
      "hi": 716.5999999999997,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 1195.6399999999921,
      "lo": 1195.6399999999921,
      "hi": 1195.6399999999921,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.0001631978021978021,
      "lo": 0.0001631978021978021,
      "hi": 0.0001631978021978021,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0.0001631978021978021,
      "lo": 0.0001631978021978021,
      "hi": 0.0001631978021978021,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=IDR",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 325.5,
      "lo": 325.5,
      "hi": 325.5,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 694.25,
      "lo": 694.25,
      "hi": 694.25,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 1136.980000000001,
      "lo": 1136.980000000001,
      "hi": 1136.980000000001,
      "n": 120,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0.0001375499999999998,
      "lo": 0.0001375499999999998,
      "hi": 0.0001375499999999998,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0.0001375499999999998,
      "lo": 0.0001375499999999998,
      "hi": 0.0001375499999999998,
      "n": 120,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "currency=USD",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.031019166418703486,
      "n": 120,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 365,
      "lo": 365,
      "hi": 365,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 739,
      "lo": 739,
      "hi": 739,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 1331.4000000000042,
      "lo": 1331.4000000000042,
      "hi": 1331.4000000000042,
      "n": 441,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.00015817687074829944,
      "lo": 0.00015817687074829944,
      "hi": 0.00015817687074829944,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0.00015817687074829944,
      "lo": 0.00015817687074829944,
      "hi": 0.00015817687074829944,
      "n": 441,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "docType=receipt",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.008635568345805946,
      "n": 441,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 355,
      "lo": 355,
      "hi": 355,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 426.15,
      "lo": 426.15,
      "hi": 426.15,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 426.83,
      "lo": 426.83,
      "hi": 426.83,
      "n": 18,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 0.00016183333333333333,
      "lo": 0.00016183333333333333,
      "hi": 0.00016183333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 0.00016183333333333333,
      "lo": 0.00016183333333333333,
      "hi": 0.00016183333333333333,
      "n": 18,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=items+total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.17587922364665765,
      "n": 18,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 378,
      "lo": 378,
      "hi": 378,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 594.5,
      "lo": 594.5,
      "hi": 594.5,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 746.4499999999999,
      "lo": 746.4499999999999,
      "hi": 746.4499999999999,
      "n": 36,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0.0001800833333333334,
      "lo": 0.0001800833333333334,
      "hi": 0.0001800833333333334,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0.0001800833333333334,
      "lo": 0.0001800833333333334,
      "hi": 0.0001800833333333334,
      "n": 36,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "editedField=total",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.09641862859446367,
      "n": 36,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 336,
      "lo": 336,
      "hi": 336,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 695.5,
      "lo": 695.5,
      "hi": 695.5,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 1068.2799999999984,
      "lo": 1068.2799999999984,
      "hi": 1068.2799999999984,
      "n": 135,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.00014264444444444434,
      "lo": 0.00014264444444444434,
      "hi": 0.00014264444444444434,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0.00014264444444444434,
      "lo": 0.00014264444444444434,
      "hi": 0.00014264444444444434,
      "n": 135,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=en",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.02766795201752491,
      "n": 135,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 365,
      "lo": 365,
      "hi": 365,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 716.5999999999997,
      "lo": 716.5999999999997,
      "hi": 716.5999999999997,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 1195.6399999999921,
      "lo": 1195.6399999999921,
      "hi": 1195.6399999999921,
      "n": 273,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.0001631978021978021,
      "lo": 0.0001631978021978021,
      "hi": 0.0001631978021978021,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0.0001631978021978021,
      "lo": 0.0001631978021978021,
      "hi": 0.0001631978021978021,
      "n": 273,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "lang=id",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.013876024339194723,
      "n": 273,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 405,
      "lo": 405,
      "hi": 405,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 1046.1499999999994,
      "lo": 1046.1499999999994,
      "hi": 1046.1499999999994,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 1473.0900000000004,
      "lo": 1473.0900000000004,
      "hi": 1473.0900000000004,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0.00018489999999999991,
      "lo": 0.00018489999999999991,
      "hi": 0.00018489999999999991,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0.00018489999999999991,
      "lo": 0.00018489999999999991,
      "hi": 0.00018489999999999991,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=explicit_amounts",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 412.5,
      "lo": 412.5,
      "hi": 412.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 753.55,
      "lo": 753.55,
      "hi": 753.55,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 1070.6600000000003,
      "lo": 1070.6600000000003,
      "hi": 1070.6600000000003,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0.0001848999999999999,
      "lo": 0.0001848999999999999,
      "hi": 0.0001848999999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0.0001848999999999999,
      "lo": 0.0001848999999999999,
      "hi": 0.0001848999999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "prompt=free",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 401.5,
      "lo": 401.5,
      "hi": 401.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 739.75,
      "lo": 739.75,
      "hi": 739.75,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 1368.9199999999987,
      "lo": 1368.9199999999987,
      "hi": 1368.9199999999987,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0.00018130000000000007,
      "lo": 0.00018130000000000007,
      "hi": 0.00018130000000000007,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0.00018130000000000007,
      "lo": 0.00018130000000000007,
      "hi": 0.00018130000000000007,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=intact",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 351,
      "lo": 351,
      "hi": 351,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 710.1,
      "lo": 710.1,
      "hi": 710.1,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 1351.0899999999963,
      "lo": 1351.0899999999963,
      "hi": 1351.0899999999963,
      "n": 330,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0.00015072727272727273,
      "lo": 0.00015072727272727273,
      "hi": 0.00015072727272727273,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0.00015072727272727273,
      "lo": 0.00015072727272727273,
      "hi": 0.00015072727272727273,
      "n": 330,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.011506835712569085,
      "n": 330,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 373.5,
      "lo": 373.5,
      "hi": 373.5,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 753.25,
      "lo": 753.25,
      "hi": 753.25,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 1019.1299999999993,
      "lo": 1019.1299999999993,
      "hi": 1019.1299999999993,
      "n": 60,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0.0001788000000000001,
      "lo": 0.0001788000000000001,
      "hi": 0.0001788000000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0.0001788000000000001,
      "lo": 0.0001788000000000001,
      "hi": 0.0001788000000000001,
      "n": 60,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "provenance=stripped",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.06017185214208986,
      "n": 60,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 367.5,
      "lo": 367.5,
      "hi": 367.5,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 748.6499999999995,
      "lo": 748.6499999999995,
      "hi": 748.6499999999995,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 1325.239999999997,
      "lo": 1325.239999999997,
      "hi": 1325.239999999997,
      "n": 228,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.00016180263157894754,
      "lo": 0.00016180263157894754,
      "hi": 0.00016180263157894754,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0.00016180263157894754,
      "lo": 0.00016180263157894754,
      "hi": 0.00016180263157894754,
      "n": 228,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=cord",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.01656933509750344,
      "n": 228,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 365,
      "lo": 365,
      "hi": 365,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 602.5999999999998,
      "lo": 602.5999999999998,
      "hi": 602.5999999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 726.9199999999998,
      "lo": 726.9199999999998,
      "hi": 726.9199999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0.0001606,
      "lo": 0.0001606,
      "hi": 0.0001606,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0.0001606,
      "lo": 0.0001606,
      "hi": 0.0001606,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 377,
      "lo": 377,
      "hi": 377,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 532.4999999999997,
      "lo": 532.4999999999997,
      "hi": 532.4999999999997,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 697.6999999999998,
      "lo": 697.6999999999998,
      "hi": 697.6999999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 0.00021039999999999994,
      "lo": 0.00021039999999999994,
      "hi": 0.00021039999999999994,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 0.00021039999999999994,
      "lo": 0.00021039999999999994,
      "hi": 0.00021039999999999994,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:gpt-image-2.5-sunburst",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 359.5,
      "lo": 359.5,
      "hi": 359.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 510.94999999999976,
      "lo": 510.94999999999976,
      "hi": 510.94999999999976,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 674.5400000000002,
      "lo": 674.5400000000002,
      "hi": 674.5400000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0.00016490000000000002,
      "lo": 0.00016490000000000002,
      "hi": 0.00016490000000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0.00016490000000000002,
      "lo": 0.00016490000000000002,
      "hi": 0.00016490000000000002,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=edited:sharp-patch",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 310.5,
      "lo": 310.5,
      "hi": 310.5,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 637.6499999999985,
      "lo": 637.6499999999985,
      "hi": 637.6499999999985,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 1219.4199999999978,
      "lo": 1219.4199999999978,
      "hi": 1219.4199999999978,
      "n": 102,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0.000125970588235294,
      "lo": 0.000125970588235294,
      "hi": 0.000125970588235294,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0.000125970588235294,
      "lo": 0.000125970588235294,
      "hi": 0.000125970588235294,
      "n": 102,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=expressexpense",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03629446214646319,
      "n": 102,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 383,
      "lo": 383,
      "hi": 383,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 1169.85,
      "lo": 1169.85,
      "hi": 1169.85,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 1492.5200000000004,
      "lo": 1492.5200000000004,
      "hi": 1492.5200000000004,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0.00016179999999999995,
      "lo": 0.00016179999999999995,
      "hi": 0.00016179999999999995,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0.00016179999999999995,
      "lo": 0.00016179999999999995,
      "hi": 0.00016179999999999995,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gemini-nano-banana-2.1",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 419.5,
      "lo": 419.5,
      "hi": 419.5,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 709.7999999999997,
      "lo": 709.7999999999997,
      "hi": 709.7999999999997,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 885.7700000000002,
      "lo": 885.7700000000002,
      "hi": 885.7700000000002,
      "n": 30,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0.0002079999999999999,
      "lo": 0.0002079999999999999,
      "hi": 0.0002079999999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0.0002079999999999999,
      "lo": 0.0002079999999999999,
      "hi": 0.0002079999999999999,
      "n": 30,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "source=generated:gpt-image-2.5-flare",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.11351339317396876,
      "n": 30,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 369,
      "lo": 369,
      "hi": 369,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 727.799999999999,
      "lo": 727.799999999999,
      "hi": 727.799999999999,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 914.0400000000001,
      "lo": 914.0400000000001,
      "hi": 914.0400000000001,
      "n": 117,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0.0001694615384615385,
      "lo": 0.0001694615384615385,
      "hi": 0.0001694615384615385,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0.0001694615384615385,
      "lo": 0.0001694615384615385,
      "hi": 0.0001694615384615385,
      "n": 117,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=exclusive",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03178924566273338,
      "n": 117,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 356,
      "lo": 356,
      "hi": 356,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 737.0999999999999,
      "lo": 737.0999999999999,
      "hi": 737.0999999999999,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 1404.2999999999997,
      "lo": 1404.2999999999997,
      "hi": 1404.2999999999997,
      "n": 99,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0.0001526060606060607,
      "lo": 0.0001526060606060607,
      "hi": 0.0001526060606060607,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0.0001526060606060607,
      "lo": 0.0001526060606060607,
      "hi": 0.0001526060606060607,
      "n": 99,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "taxMode=none",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.03735321206782739,
      "n": 99,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 323,
      "lo": 323,
      "hi": 323,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 462.2999999999999,
      "lo": 462.2999999999999,
      "hi": 462.2999999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 524.4599999999999,
      "lo": 524.4599999999999,
      "hi": 524.4599999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 0.00014720000000000003,
      "lo": 0.00014720000000000003,
      "hi": 0.00014720000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 0.00014720000000000003,
      "lo": 0.00014720000000000003,
      "hi": 0.00014720000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=crop",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 330,
      "lo": 330,
      "hi": 330,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 796.099999999999,
      "lo": 796.099999999999,
      "hi": 796.099999999999,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 1281.6199999999994,
      "lo": 1281.6199999999994,
      "hi": 1281.6199999999994,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 0.00016100000000000004,
      "lo": 0.00016100000000000004,
      "hi": 0.00016100000000000004,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 0.00016100000000000004,
      "lo": 0.00016100000000000004,
      "hi": 0.00016100000000000004,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=jpeg",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 312,
      "lo": 312,
      "hi": 312,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 431.2,
      "lo": 431.2,
      "hi": 431.2,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 444.64,
      "lo": 444.64,
      "hi": 444.64,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 0.00016059999999999997,
      "lo": 0.00016059999999999997,
      "hi": 0.00016059999999999997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 0.00016059999999999997,
      "lo": 0.00016059999999999997,
      "hi": 0.00016059999999999997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rephoto",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 252,
      "lo": 252,
      "hi": 252,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 349.69999999999993,
      "lo": 349.69999999999993,
      "hi": 349.69999999999993,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 399.53999999999996,
      "lo": 399.53999999999996,
      "hi": 399.53999999999996,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 0.00007879999999999999,
      "lo": 0.00007879999999999999,
      "hi": 0.00007879999999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 0.00007879999999999999,
      "lo": 0.00007879999999999999,
      "hi": 0.00007879999999999999,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=resize",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 360,
      "lo": 360,
      "hi": 360,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 579.4999999999995,
      "lo": 579.4999999999995,
      "hi": 579.4999999999995,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 817.4999999999998,
      "lo": 817.4999999999998,
      "hi": 817.4999999999998,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 0.00016719999999999997,
      "lo": 0.00016719999999999997,
      "hi": 0.00016719999999999997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 0.00016719999999999997,
      "lo": 0.00016719999999999997,
      "hi": 0.00016719999999999997,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=rotate",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p50_ms",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 440,
      "lo": 440,
      "hi": 440,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p95_ms",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 1008.7999999999993,
      "lo": 1008.7999999999993,
      "hi": 1008.7999999999993,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "latency_p99_ms",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 1376.1599999999996,
      "lo": 1376.1599999999996,
      "hi": 1376.1599999999996,
      "n": 15,
      "unit": "ms",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0.00015500000000000003,
      "lo": 0.00015500000000000003,
      "hi": 0.00015500000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "model_cost_per_check_usd",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0.00015500000000000003,
      "lo": 0.00015500000000000003,
      "hi": 0.00015500000000000003,
      "n": 15,
      "unit": "usd",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "error_rate",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "refusal_rate",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    },
    {
      "task": "operations",
      "metric": "timeout_rate",
      "stratum": "transform=screenshot",
      "target": "decision:gpt-6-luna",
      "value": 0,
      "lo": 0,
      "hi": 0.20388330103584862,
      "n": 15,
      "unit": "rate",
      "higherIsBetter": false
    }
  ],
  "findings": {
    "losses": [
      "operations, cost per check to the caller: Receipt Audit (Claude Haiku 5.5) $0.016 (n=450) vs Claude Haiku 5.5, plain prompt $0.00081 (n=450).",
      "operations, cost per check to the caller: Receipt Audit (Claude Haiku 5.5) $0.016 (n=450) vs Gemini 3.8 Flash, plain prompt $0.0043 (n=450).",
      "operations, cost per check to the caller: Receipt Audit (Claude Haiku 5.5) $0.016 (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake $0.00016 (n=450).",
      "operations, cost per check to the caller: Receipt Audit on Gemini 3.8 Flash $0.016 (n=450) vs Claude Haiku 5.5, plain prompt $0.00081 (n=450).",
      "operations, cost per check to the caller: Receipt Audit on Gemini 3.8 Flash $0.016 (n=450) vs Gemini 3.8 Flash, plain prompt $0.0043 (n=450).",
      "operations, cost per check to the caller: Receipt Audit on Gemini 3.8 Flash $0.016 (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake $0.00016 (n=450).",
      "operations, model cost per check: Receipt Audit (Claude Haiku 5.5) $0.0011 (n=450) vs Claude Haiku 5.5, plain prompt $0.00081 (n=450).",
      "operations, model cost per check: Receipt Audit (Claude Haiku 5.5) $0.0011 (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake $0.00016 (n=450).",
      "operations, model cost per check: Receipt Audit on Gemini 3.8 Flash $0.0044 (n=450) vs Claude Haiku 5.5, plain prompt $0.00081 (n=450).",
      "operations, model cost per check: Receipt Audit on Gemini 3.8 Flash $0.0044 (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake $0.00016 (n=450).",
      "operations, median latency: Receipt Audit (Claude Haiku 5.5) 6,299 ms (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake 365 ms (n=450).",
      "operations, median latency: Receipt Audit on Gemini 3.8 Flash 4,518 ms (n=450) vs OpenAI decision model (gpt-6-luna), asked if it's fake 365 ms (n=450)."
    ],
    "wins": [
      "extraction, documents with every labelled field right: Receipt Audit (Claude Haiku 5.5) 94.4% [88.8%–97.3%] (n=120); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "extraction, documents with every labelled field right: Receipt Audit on Gemini 3.8 Flash 96.7% [91.7%–98.7%] (n=120); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "extraction, total read right: Receipt Audit (Claude Haiku 5.5) 97.2% [92.5%–99%] (n=120); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "extraction, total read right: Receipt Audit on Gemini 3.8 Flash 98.3% [94.1%–99.5%] (n=120); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "extraction, line-item F1: Receipt Audit (Claude Haiku 5.5) 93.1% [87.4%–97.1%] (n=98); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "extraction, line-item F1: Receipt Audit on Gemini 3.8 Flash 94.1% [88.9%–98.3%] (n=98); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit (Claude Haiku 5.5) 86.7% [70.3%–94.7%] (n=30); Claude Haiku 5.5, plain prompt can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit (Claude Haiku 5.5) 86.7% [70.3%–94.7%] (n=30); Gemini 3.8 Flash, plain prompt can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit (Claude Haiku 5.5) 86.7% [70.3%–94.7%] (n=30); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit on Gemini 3.8 Flash 85.6% [69%–94%] (n=30); Claude Haiku 5.5, plain prompt can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit on Gemini 3.8 Flash 85.6% [69%–94%] (n=30); Gemini 3.8 Flash, plain prompt can't do this task.",
      "duplicates, resubmitted copies recognised: Receipt Audit on Gemini 3.8 Flash 85.6% [69%–94%] (n=30); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "decisions, calibration error of the risk score: Receipt Audit on Gemini 3.8 Flash 2.3% [1.9%–7%] (n=120) vs Claude Haiku 5.5, plain prompt 10.7% [8.5%–13.9%] (n=120).",
      "proof, fakes caught with verifiable evidence: Receipt Audit (Claude Haiku 5.5) 46.7% [32.2%–61.7%] (n=40); Claude Haiku 5.5, plain prompt can't do this task.",
      "proof, fakes caught with verifiable evidence: Receipt Audit (Claude Haiku 5.5) 46.7% [32.2%–61.7%] (n=40); Gemini 3.8 Flash, plain prompt can't do this task.",
      "proof, fakes caught with verifiable evidence: Receipt Audit (Claude Haiku 5.5) 46.7% [32.2%–61.7%] (n=40); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "proof, fakes caught with verifiable evidence: Receipt Audit on Gemini 3.8 Flash 60% [44.6%–73.7%] (n=40); Claude Haiku 5.5, plain prompt can't do this task.",
      "proof, fakes caught with verifiable evidence: Receipt Audit on Gemini 3.8 Flash 60% [44.6%–73.7%] (n=40); Gemini 3.8 Flash, plain prompt can't do this task.",
      "proof, fakes caught with verifiable evidence: Receipt Audit on Gemini 3.8 Flash 60% [44.6%–73.7%] (n=40); OpenAI decision model (gpt-6-luna), asked if it's fake can't do this task.",
      "operations, model cost per check: Receipt Audit (Claude Haiku 5.5) $0.0011 (n=450) vs Gemini 3.8 Flash, plain prompt $0.0043 (n=450).",
      "operations, median latency: Receipt Audit on Gemini 3.8 Flash 4,518 ms (n=450) vs Claude Haiku 5.5, plain prompt 7,213 ms (n=450).",
      "operations, median latency: Receipt Audit on Gemini 3.8 Flash 4,518 ms (n=450) vs Gemini 3.8 Flash, plain prompt 6,719 ms (n=450)."
    ],
    "ties": [
      "extraction, documents with every labelled field right: Receipt Audit (Claude Haiku 5.5) 94.4% [88.8%–97.3%] (n=120) vs Claude Haiku 5.5, plain prompt 88.6% [81.7%–93.1%] (n=120).",
      "extraction, documents with every labelled field right: Receipt Audit (Claude Haiku 5.5) 94.4% [88.8%–97.3%] (n=120) vs Gemini 3.8 Flash, plain prompt 92.8% [86.7%–96.2%] (n=120).",
      "extraction, documents with every labelled field right: Receipt Audit on Gemini 3.8 Flash 96.7% [91.7%–98.7%] (n=120) vs Claude Haiku 5.5, plain prompt 88.6% [81.7%–93.1%] (n=120).",
      "extraction, documents with every labelled field right: Receipt Audit on Gemini 3.8 Flash 96.7% [91.7%–98.7%] (n=120) vs Gemini 3.8 Flash, plain prompt 92.8% [86.7%–96.2%] (n=120).",
      "extraction, total read right: Receipt Audit (Claude Haiku 5.5) 97.2% [92.5%–99%] (n=120) vs Claude Haiku 5.5, plain prompt 97.2% [92.5%–99%] (n=120).",
      "extraction, total read right: Receipt Audit (Claude Haiku 5.5) 97.2% [92.5%–99%] (n=120) vs Gemini 3.8 Flash, plain prompt 99.2% [95.4%–99.9%] (n=120).",
      "extraction, total read right: Receipt Audit on Gemini 3.8 Flash 98.3% [94.1%–99.5%] (n=120) vs Claude Haiku 5.5, plain prompt 97.2% [92.5%–99%] (n=120).",
      "extraction, total read right: Receipt Audit on Gemini 3.8 Flash 98.3% [94.1%–99.5%] (n=120) vs Gemini 3.8 Flash, plain prompt 99.2% [95.4%–99.9%] (n=120).",
      "extraction, line-item F1: Receipt Audit (Claude Haiku 5.5) 93.1% [87.4%–97.1%] (n=98) vs Claude Haiku 5.5, plain prompt 91.7% [86.6%–95.4%] (n=98).",
      "extraction, line-item F1: Receipt Audit (Claude Haiku 5.5) 93.1% [87.4%–97.1%] (n=98) vs Gemini 3.8 Flash, plain prompt 96.7% [92.4%–99.3%] (n=98).",
      "extraction, line-item F1: Receipt Audit on Gemini 3.8 Flash 94.1% [88.9%–98.3%] (n=98) vs Claude Haiku 5.5, plain prompt 91.7% [86.6%–95.4%] (n=98).",
      "extraction, line-item F1: Receipt Audit on Gemini 3.8 Flash 94.1% [88.9%–98.3%] (n=98) vs Gemini 3.8 Flash, plain prompt 96.7% [92.4%–99.3%] (n=98).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit (Claude Haiku 5.5) 0.938 [0.881–0.984] (n=120) vs Claude Haiku 5.5, plain prompt 0.903 [0.832–0.961] (n=120).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit (Claude Haiku 5.5) 0.938 [0.881–0.984] (n=120) vs Gemini 3.8 Flash, plain prompt 0.939 [0.877–0.989] (n=120).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit (Claude Haiku 5.5) 0.938 [0.881–0.984] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 0.911 [0.840–0.966] (n=120).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit on Gemini 3.8 Flash 0.944 [0.889–0.987] (n=120) vs Claude Haiku 5.5, plain prompt 0.903 [0.832–0.961] (n=120).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit on Gemini 3.8 Flash 0.944 [0.889–0.987] (n=120) vs Gemini 3.8 Flash, plain prompt 0.939 [0.877–0.989] (n=120).",
      "authenticity, ROC AUC, genuine vs fake: Receipt Audit on Gemini 3.8 Flash 0.944 [0.889–0.987] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 0.911 [0.840–0.966] (n=120).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit (Claude Haiku 5.5) 85% [72.5%–97.5%] (n=40) vs Claude Haiku 5.5, plain prompt 70% [52.5%–87.5%] (n=40).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit (Claude Haiku 5.5) 85% [72.5%–97.5%] (n=40) vs Gemini 3.8 Flash, plain prompt 82.5% [70%–92.5%] (n=40).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit (Claude Haiku 5.5) 85% [72.5%–97.5%] (n=40) vs OpenAI decision model (gpt-6-luna), asked if it's fake 75% [62.5%–90%] (n=40).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit on Gemini 3.8 Flash 90% [80%–97.5%] (n=40) vs Claude Haiku 5.5, plain prompt 70% [52.5%–87.5%] (n=40).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit on Gemini 3.8 Flash 90% [80%–97.5%] (n=40) vs Gemini 3.8 Flash, plain prompt 82.5% [70%–92.5%] (n=40).",
      "authenticity, fakes caught at a 5% false-flag rate: Receipt Audit on Gemini 3.8 Flash 90% [80%–97.5%] (n=40) vs OpenAI decision model (gpt-6-luna), asked if it's fake 75% [62.5%–90%] (n=40).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit (Claude Haiku 5.5) 76.9% [61.5%–96.2%] (n=26) vs Claude Haiku 5.5, plain prompt 53.8% [30.8%–76.9%] (n=26).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit (Claude Haiku 5.5) 76.9% [61.5%–96.2%] (n=26) vs Gemini 3.8 Flash, plain prompt 73.1% [53.8%–88.5%] (n=26).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit (Claude Haiku 5.5) 76.9% [61.5%–96.2%] (n=26) vs OpenAI decision model (gpt-6-luna), asked if it's fake 80.8% [65.4%–96.2%] (n=26).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit on Gemini 3.8 Flash 84.6% [69.2%–96.2%] (n=26) vs Claude Haiku 5.5, plain prompt 53.8% [30.8%–76.9%] (n=26).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit on Gemini 3.8 Flash 84.6% [69.2%–96.2%] (n=26) vs Gemini 3.8 Flash, plain prompt 73.1% [53.8%–88.5%] (n=26).",
      "authenticity, fakes whose amounts add up, caught at a 5% false-flag rate (arithmetic=consistent): Receipt Audit on Gemini 3.8 Flash 84.6% [69.2%–96.2%] (n=26) vs OpenAI decision model (gpt-6-luna), asked if it's fake 80.8% [65.4%–96.2%] (n=26).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) 0% [0%–4.6%] (n=80) vs Claude Haiku 5.5, plain prompt 0.8% [0.1%–6.1%] (n=80).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) 0% [0%–4.6%] (n=80) vs Gemini 3.8 Flash, plain prompt 1.3% [0.2%–6.7%] (n=80).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) 0% [0%–4.6%] (n=80) vs OpenAI decision model (gpt-6-luna), asked if it's fake 0% [0%–4.6%] (n=80).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash 1.7% [0.4%–7.4%] (n=80) vs Claude Haiku 5.5, plain prompt 0.8% [0.1%–6.1%] (n=80).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash 1.7% [0.4%–7.4%] (n=80) vs Gemini 3.8 Flash, plain prompt 1.3% [0.2%–6.7%] (n=80).",
      "authenticity, genuine receipts flagged at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash 1.7% [0.4%–7.4%] (n=80) vs OpenAI decision model (gpt-6-luna), asked if it's fake 0% [0%–4.6%] (n=80).",
      "decisions, same decision in every repetition: Receipt Audit (Claude Haiku 5.5) 95.8% [90.6%–98.2%] (n=120) vs Claude Haiku 5.5, plain prompt 90% [83.3%–94.2%] (n=120).",
      "decisions, same decision in every repetition: Receipt Audit (Claude Haiku 5.5) 95.8% [90.6%–98.2%] (n=120) vs Gemini 3.8 Flash, plain prompt 96.7% [91.7%–98.7%] (n=120).",
      "decisions, same decision in every repetition: Receipt Audit (Claude Haiku 5.5) 95.8% [90.6%–98.2%] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 100% [96.9%–100%] (n=120).",
      "decisions, same decision in every repetition: Receipt Audit on Gemini 3.8 Flash 97.5% [92.9%–99.1%] (n=120) vs Claude Haiku 5.5, plain prompt 90% [83.3%–94.2%] (n=120).",
      "decisions, same decision in every repetition: Receipt Audit on Gemini 3.8 Flash 97.5% [92.9%–99.1%] (n=120) vs Gemini 3.8 Flash, plain prompt 96.7% [91.7%–98.7%] (n=120).",
      "decisions, same decision in every repetition: Receipt Audit on Gemini 3.8 Flash 97.5% [92.9%–99.1%] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 100% [96.9%–100%] (n=120).",
      "decisions, calibration error of the risk score: Receipt Audit (Claude Haiku 5.5) 8.5% [6.1%–12.3%] (n=120) vs Claude Haiku 5.5, plain prompt 10.7% [8.5%–13.9%] (n=120).",
      "decisions, calibration error of the risk score: Receipt Audit (Claude Haiku 5.5) 8.5% [6.1%–12.3%] (n=120) vs Gemini 3.8 Flash, plain prompt 3.3% [1.7%–7.2%] (n=120).",
      "decisions, calibration error of the risk score: Receipt Audit (Claude Haiku 5.5) 8.5% [6.1%–12.3%] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 6.5% [4.4%–12.2%] (n=120).",
      "decisions, calibration error of the risk score: Receipt Audit on Gemini 3.8 Flash 2.3% [1.9%–7%] (n=120) vs Gemini 3.8 Flash, plain prompt 3.3% [1.7%–7.2%] (n=120).",
      "decisions, calibration error of the risk score: Receipt Audit on Gemini 3.8 Flash 2.3% [1.9%–7%] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake 6.5% [4.4%–12.2%] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) $400.000 [$258.333–$541.875] (n=120) vs Claude Haiku 5.5, plain prompt $631.917 [$416.667–$883.958] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) $400.000 [$258.333–$541.875] (n=120) vs Gemini 3.8 Flash, plain prompt $402.042 [$145.039–$764.494] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit (Claude Haiku 5.5) $400.000 [$258.333–$541.875] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake $350.000 [$200.000–$500.000] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash $447.167 [$108.333–$1016.535] (n=120) vs Claude Haiku 5.5, plain prompt $631.917 [$416.667–$883.958] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash $447.167 [$108.333–$1016.535] (n=120) vs Gemini 3.8 Flash, plain prompt $402.042 [$145.039–$764.494] (n=120).",
      "decisions, expected cost of decisions per 1,000 receipts at risk ≥ 0.5: Receipt Audit on Gemini 3.8 Flash $447.167 [$108.333–$1016.535] (n=120) vs OpenAI decision model (gpt-6-luna), asked if it's fake $350.000 [$200.000–$500.000] (n=120).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit (Claude Haiku 5.5) 30% [14.5%–51.9%] (n=20) vs Claude Haiku 5.5, plain prompt 58.3% [37.2%–76.8%] (n=20).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit (Claude Haiku 5.5) 30% [14.5%–51.9%] (n=20) vs Gemini 3.8 Flash, plain prompt 61.7% [40.2%–79.4%] (n=20).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit (Claude Haiku 5.5) 30% [14.5%–51.9%] (n=20) vs OpenAI decision model (gpt-6-luna), asked if it's fake 40% [21.9%–61.3%] (n=20).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit on Gemini 3.8 Flash 63.3% [41.7%–80.6%] (n=20) vs Claude Haiku 5.5, plain prompt 58.3% [37.2%–76.8%] (n=20).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit on Gemini 3.8 Flash 63.3% [41.7%–80.6%] (n=20) vs Gemini 3.8 Flash, plain prompt 61.7% [40.2%–79.4%] (n=20).",
      "evidence, edited fakes flagged with the edited field named: Receipt Audit on Gemini 3.8 Flash 63.3% [41.7%–80.6%] (n=20) vs OpenAI decision model (gpt-6-luna), asked if it's fake 40% [21.9%–61.3%] (n=20).",
      "operations, model cost per check: Receipt Audit on Gemini 3.8 Flash $0.0044 (n=450) vs Gemini 3.8 Flash, plain prompt $0.0043 (n=450).",
      "operations, median latency: Receipt Audit (Claude Haiku 5.5) 6,299 ms (n=450) vs Claude Haiku 5.5, plain prompt 7,213 ms (n=450).",
      "operations, median latency: Receipt Audit (Claude Haiku 5.5) 6,299 ms (n=450) vs Gemini 3.8 Flash, plain prompt 6,719 ms (n=450)."
    ]
  },
  "datasets": [
    {
      "name": "CORD v2 (naver-clova-ix/cord-v2)",
      "license": "CC-BY-4.0",
      "url": "https://huggingface.co/datasets/naver-clova-ix/cord-v2",
      "cases": 76
    },
    {
      "name": "ExpressExpense SRD (sample receipt dataset)",
      "license": "MIT",
      "url": "https://expressexpense.com/blog/free-receipt-images-ocr-machine-learning-dataset/",
      "cases": 34
    },
    {
      "name": "RA-Bench generated fakes",
      "license": "CC-BY-4.0",
      "url": null,
      "cases": 20
    },
    {
      "name": "RA-Bench fakes from CORD v2",
      "license": "CC-BY-4.0",
      "url": null,
      "cases": 14
    },
    {
      "name": "RA-Bench fakes from ExpressExpense SRD",
      "license": "MIT",
      "url": null,
      "cases": 6
    }
  ],
  "limitations": [
    "Preliminary: small samples, so intervals are wide. Read the intervals, not the point estimates.",
    "Public cases come from datasets that are probably in the models' training data.",
    "The DIY baselines use one plain prompt; a stronger prompt would do better. A strong-prompt baseline comes later.",
    "DIY targets have no memory across checks and no deterministic checks: those tasks are marked not supported, not scored as zero.",
    "Our engine runs in process, so its latency has no network hop to receiptaudit.com; DIY latency includes the round trip to the model.",
    "Costs are model tokens at list price on the price date; no infrastructure cost is included.",
    "When the Claude baseline's config says transport claude-code-cli, it ran through Claude Code's CLI on a subscription: thinking is adaptive and can't be fully turned off, its cost is what the same tokens cost through the API, and its latency includes the CLI's start-up.",
    "The duplicate network starts empty for each repetition: it only knows the run's own earlier cases, not production traffic."
  ],
  "examples": [],
  "methodology": "RA-Bench 0.2.0. These definitions, thresholds and exclusions were written before the run and change only in a new suite version.\n\nWho and what. Receipt Audit runs this benchmark on its own product. Every target gets the same cases, the same files, the same timeout (180 s per attempt by default) and the same scoring; only the target changes. A target is our engine (\"self\", in process, with the duplicate network on an empty in-memory backend that lives for one repetition) or an agent's own model asked with one plain prompt (\"diy\"). DIY targets get the same file preparation as our engine (upright, at most 3072 px and 3.75 MB) so that API size limits don't count against them. Transport failures (network errors, HTTP 408, 429 and 5xx) are retried up to 3 times with exponential backoff; nothing else is retried, and a wrong answer is a result. Models are pinned: no fallback to another model.\n\nUnits and intervals. A case run several times counts once, with the share of its runs that succeeded. Rates carry a 95% Wilson interval over cases (operations rates: over calls). ROC AUC, detection at a fixed false-flag rate and line-item F1 carry a 95% percentile bootstrap interval (2000 resamples, genuine receipts and fakes resampled separately, seeded per row so a re-scored run gives the same interval). Latency and cost have no interval. Every row reports n. Rows are given for all cases and for every stratum (a \"dimension=value\" tag on the case, or a label-derived group) with at least 10 cases.\n\nWhich cases count. Extraction, checks and authenticity use base cases only: cases that aren't a transform of another case (no lineage). Duplicate transforms count only for the duplicates task. Operations use every call.\n\nExtraction. Scored only on fields the label covers; a field labelled null means the document doesn't show it, and only a null answer is right. Merchant name: lowercase, accents stripped, \"&\" read as \"and\", punctuation dropped, then a match at edit-distance similarity ≥ 0.9. Date: the exact ISO date (YYYY-MM-DD). Subtotal, tax total and total: within 0.01. Currency: the exact ISO 4217 code. Receipt number: equal after keeping only letters and digits, lowercased. \"All fields right\": every labelled scalar field of the document right in that run (line items excluded). Line-item F1: labelled and extracted items paired one to one when descriptions are at least 0.8 alike (same normalisation) and totals match within 0.01, micro-averaged over all documents. Valid output: the call succeeded and returned readable fields. A failed call or an unreadable answer gets every field wrong.\n\nChecks. totals_add_up is scored against the label \"the printed amounts add up\". A positive is a document whose amounts don't add up; the check flags it when it reports failed. Recall over documents that don't add up; precision over documents it flagged; false alarms over genuine documents that do add up; coverage is the share of documents where the check ran (a check that couldn't run isn't output). Targets without deterministic checks report this task as not supported.\n\nAuthenticity (genuine vs fake). Higher risk means more likely fake. ROC AUC uses each target's risk as the score; a result without a risk scores 0.5 (no opinion), ties count half. Detection at a 1%, 5% and 10% false-flag rate: the share of fakes flagged by the lowest threshold that flags at most that share of genuine receipts (ties at the threshold are all flagged). At the documented thresholds risk ≥ 0.5 and risk ≥ 0.7: the share of fakes flagged (detection) and of genuine receipts flagged (false flags); a result without a risk is never flagged. Strata from the labels: each fake class (\"class=ai_full\", …) and fakes whose amounts add up or don't (\"arithmetic=consistent\", \"arithmetic=inconsistent\"), each compared with every genuine receipt; and provenance metadata (\"provenance=intact\", \"stripped\", \"none\"), genuine receipts and fakes alike.\n\nDuplicates. Every case is submitted by a different account, so a repeat is \"someone else submitted this receipt before\". Recall: the share of duplicate transforms (re-encoded, cropped, re-shot copies of an earlier case) the target recognised as seen before; exact recall counts only matches on the identical image hash. False matches: the share of distinct receipts reported as seen before, for hard negatives (different receipts that look alike: same merchant, template or day; \"false_match_rate\") and for the other genuine receipts (\"false_match_rate_genuine\"). Originals always run before their transforms. Targets without memory across checks (DIY prompts) report the task as not supported, not as zero.\n\nDecisions. A decision is the yes/no a caller acts on: flagged when risk ≥ 0.5. Consistency: over base cases run at least twice, the share whose decision is the same in every repetition; total consistency, the share whose extracted total is the same (within 0.01, or absent) in every repetition. Calibration error: expected calibration error of risk against \"is fake\" over every run with a risk, 10 equal-width bins, weighted by run count (lower is better; it depends on this set's share of fakes). Expected cost per 1,000 receipts at risk ≥ 0.5 and ≥ 0.7: 1000 × ((1 − p) × false-flag rate × C_review + p × (1 − detection) × C_miss). Defaults, each from a cited, graded source (research/41-benchmark-cost-model.md): p = 1% of submitted receipts are fake (SAP Concur's share of receipts flagged as AI-generated; no study measures the true share, so this is the best available assumption, not a measurement); C_miss = $100, the face value of a missed fake (AppZen's mean AI-generated fake, 12 months to May 2026); C_review = $17, one false flag's review (18 minutes, GBTA, at a loaded auditor wage of $57.47/h, BLS). At these defaults one point of false-flag rate costs as much as about 17 points of detection. The sweep varies one assumption at a time over its published spread (\"assumption=…\": fakes 0.1% and 5%; a missed fake $32 and $182; a review $8 and $73). Approving everything costs 1000 × p × C_miss ($1000 at the defaults), reported as its own row: a detector above it loses money under those assumptions. The formula counts only mistakes (it treats reviewing a fake that was rightly flagged as free) and leaves out unmeasured costs: an employee's time and trust after a false flag, and repeat fraud or tax penalties after a miss. This set has far more fakes than real submissions, so the cost re-weights the measured rates by p rather than using the set's own mix.\n\nEvidence. A yes/no is easy to get from any model; a reviewer also needs to know what is wrong. Edited field named: over fakes whose label names the edited field (AI-inpainted and hand-edited fakes), the share of runs that flag the receipt and whose first 3 reasons point at that field. A reason points at a field through an explicit \"@field\" tag, a deterministic check about it (an arithmetic check points at the total and subtotal), or, for free-text reasons, a pre-registered keyword (total, amount due; subtotal; tax, VAT, GST; date; merchant, store, vendor, logo, name; item, line, quantity, price).\n\nVerifiable evidence. A finding is proof of tampering when a caller can check it without trusting a model and it shows the document was generated or changed: an AI-generator, AI source-type or receipt/invoice-generator tag read from the file, PDF modification records, a thumbnail that doesn't match the image, a photo taken before the printed purchase date, printed amounts that don't add up (with the numbers), or a duplicate-network match. Other file facts (an editor or office tool, missing camera data) and plausibility or policy checks (tax rate, the receipt's age, currency) are real but don't show tampering, and don't count. A model's visual judgement is an opinion, however confident. Fakes caught with proof: the share of fakes flagged with at least one proof among their reasons. Genuine receipts with proof: the guard against false proof (lower is better). Only targets that can return verifiable findings report this task; a plain prompt returns opinions by construction and reports it as not supported.\n\nOperations. Latency: wall time of the final attempt of each successful call, measured by the harness (p50, p95, p99). Cost per check is what the caller pays for one check: for Receipt Audit, the list price of its smallest credit pack per check; for a DIY baseline, its model tokens × the AI Gateway list price on the run's price date. Model cost per check is the model tokens × that list price for every target, our engine's own model call included (infrastructure not included). Both are averaged over successful calls. Error, refusal and timeout rates over all calls.\n\nFindings. Each of our targets is compared with each baseline on the headline rows of every task (stratum \"all\"): a win or a loss only when the 95% intervals don't overlap, otherwise a tie; for latency and cost, which have no interval, only when one value is at least 1.2 times the other. A comparison resting on fewer than 10 cases (or calls) on either side is always a tie. A task a baseline can't do by construction is reported as such, not as a score of zero.",
  "reproduce": {
    "repo": null,
    "command": "cd apps/receipt-check && nub run bench run --cases benchmarks/receipt-audit/cases.jsonl --split public --targets self,self-gemini,diy:claude-haiku-5.5,diy:gemini-3.8-flash,decision:gpt-6-luna --reps 3"
  },
  "history": [
    {
      "runId": "20261012T223232Z-63be0d",
      "ranAt": "2026-10-12T22:49:04.881Z",
      "url": "https://receiptaudit.com/benchmarks/runs/20261012T223232Z-63be0d.json"
    },
    {
      "runId": "20261012T203648Z-44a89a",
      "ranAt": "2026-10-12T20:47:27.588Z",
      "url": "https://receiptaudit.com/benchmarks/runs/20261012T203648Z-44a89a.json"
    }
  ]
}