{
  "artifactType": "golden_eval_public_summary",
  "classes": {
    "app_bug": {
      "correct": 7,
      "f1": 0.8235,
      "precision": 0.7,
      "predicted": 10,
      "recall": 1,
      "support": 7
    },
    "bad_assertion": {
      "correct": 4,
      "f1": 1,
      "precision": 1,
      "predicted": 4,
      "recall": 1,
      "support": 4
    },
    "confirmed_pass": {
      "correct": 7,
      "f1": 1,
      "precision": 1,
      "predicted": 7,
      "recall": 1,
      "support": 7
    },
    "env_fixture": {
      "correct": 3,
      "f1": 1,
      "precision": 1,
      "predicted": 3,
      "recall": 1,
      "support": 3
    },
    "false_positive": {
      "correct": 2,
      "f1": 0.5,
      "precision": 1,
      "predicted": 2,
      "recall": 0.3333,
      "support": 6
    },
    "flake": {
      "correct": 5,
      "f1": 1,
      "precision": 1,
      "predicted": 5,
      "recall": 1,
      "support": 5
    },
    "inconclusive": {
      "correct": 5,
      "f1": 0.9091,
      "precision": 0.8333,
      "predicted": 6,
      "recall": 1,
      "support": 5
    }
  },
  "description": "Public summary of the frozen judge benchmark. Labels are ground truth about each run artifact, authored from run semantics, never copied from judge output. Case bodies are not published; per-class metrics, the misses by class, and the run history are.",
  "engineCommit": "f6529f8",
  "evaluatedAt": "2026-07-29T14:57:32.792Z",
  "history": [
    {
      "accuracy": 1,
      "caseCount": 12,
      "comparable": false,
      "generatedAt": "2026-07-27T04:54:11.151Z",
      "judgeId": null,
      "macroF1": 1,
      "suiteFingerprint": "2d9664a88f4fc7fc"
    },
    {
      "accuracy": 1,
      "caseCount": 12,
      "comparable": false,
      "generatedAt": "2026-07-27T09:58:54.615Z",
      "judgeId": "offline-heuristic",
      "macroF1": 1,
      "suiteFingerprint": "2d9664a88f4fc7fc"
    },
    {
      "accuracy": 0.8378,
      "caseCount": 37,
      "comparable": false,
      "generatedAt": "2026-07-27T19:35:43.180Z",
      "judgeId": "offline-heuristic",
      "macroF1": 0.8904,
      "suiteFingerprint": "7feddeba073cb4ca"
    },
    {
      "accuracy": 0.8378,
      "caseCount": 37,
      "comparable": false,
      "generatedAt": "2026-07-27T19:36:01.400Z",
      "judgeId": "offline-heuristic",
      "macroF1": 0.8904,
      "suiteFingerprint": "7feddeba073cb4ca"
    },
    {
      "accuracy": 0.8919,
      "caseCount": 37,
      "comparable": true,
      "generatedAt": "2026-07-27T19:37:07.526Z",
      "judgeId": "offline-heuristic",
      "macroF1": 0.8904,
      "suiteFingerprint": "7453fb3027289645"
    }
  ],
  "judge": {
    "id": "offline-heuristic",
    "kind": "offline heuristic, no LLM call"
  },
  "misses": [
    {
      "confidence": 0.9,
      "expected": "false_positive",
      "got": "app_bug",
      "id": "pa_functional_regression_061da18e"
    },
    {
      "confidence": 0.9,
      "expected": "false_positive",
      "got": "app_bug",
      "id": "pa_functional_regression_22c7e763"
    },
    {
      "confidence": 0.9,
      "expected": "false_positive",
      "got": "app_bug",
      "id": "pa_functional_regression_df16e69e"
    },
    {
      "confidence": 0.5,
      "expected": "false_positive",
      "got": "inconclusive",
      "id": "pa_needs_triage_79ec441f"
    }
  ],
  "notes": [
    "Gate floors: app_bug recall and confirmed_pass recall must be 1.0 or the evaluation fails. A missed real bug or a guessed pass is the error class treated as unacceptable.",
    "false_positive recall is 0.3333: 4 artifacts that a human labelled as false positives were classified as app_bug (3), inconclusive (1) and routed to review. The miss direction is safe; the cost is triage time.",
    "History rows whose suite fingerprint differs from the current suite are earlier, smaller suites and are not comparable.",
    "This file is generated from the engine evaluation output by qa-forge-core CI on every merge to main; evaluatedAt is the freshness of the underlying evaluation."
  ],
  "result": {
    "accuracy": 0.8919,
    "caseCount": 37,
    "correct": 33,
    "gateFloors": {
      "app_bug": 1,
      "confirmed_pass": 1
    },
    "macroF1": 0.8904,
    "verdict": "pass"
  },
  "schemaVersion": 1,
  "suite": {
    "caseCount": 37,
    "coverage": {
      "app_bug": 7,
      "bad_assertion": 4,
      "confirmed_pass": 7,
      "env_fixture": 3,
      "false_positive": 6,
      "flake": 5,
      "inconclusive": 5
    },
    "fingerprint": "7453fb3027289645",
    "frozenAt": "2026-07-27",
    "id": "run-artifact-judge-v1"
  }
}