{
  "status": "PASS",
  "corpus": "mvp0-2026-09-28.2",
  "corpus_kind": "synthetic deterministic fixtures; not live Dovren or customer performance",
  "source": {
    "head_commit": "d6d696a2ec20e414b26d2d86f8a63296a4d31803",
    "working_tree": "clean"
  },
  "runtime": {
    "node": "v22.23.2",
    "platform": "linux-x64"
  },
  "postgres": "PostgreSQL 18.4 on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0, 64-bit",
  "evaluated_cases": 100,
  "status_counts": {
    "VERIFIED_PASS": 40,
    "VERIFIED_FAIL": 25,
    "INSUFFICIENT_EVIDENCE": 20,
    "HUMAN_REVIEW": 8,
    "NOT_APPLICABLE": 7
  },
  "metrics": {
    "verified_task_success_rate": {
      "numerator": 40,
      "denominator": 93,
      "value": 0.43010752688172044
    },
    "false_verified_rate": {
      "numerator": 0,
      "denominator": 25,
      "value": 0
    },
    "critical_failure_recall": {
      "detected": 25,
      "abstained": 4,
      "denominator": 29,
      "value": 1
    },
    "oracle_coverage": {
      "numerator": 65,
      "denominator": 93,
      "value": 0.6989247311827957
    },
    "double_labeled_human_agreement": {
      "numerator": null,
      "denominator": null,
      "value": null
    },
    "hosted_cost_per_run": null
  },
  "controls": {
    "claim_only_false_verified_on_known_negatives": 25,
    "langfuse_oss_baseline": "not_run"
  },
  "latency_ms": {
    "p50": 13.249786999999742,
    "p95": 23.512234000000035,
    "sample_count": 100
  },
  "reproducibility": {
    "repeated_evaluations_preserved_all_100_verdicts": true
  },
  "traceability": {
    "verdicts_with_criteria_rubric_and_verifier_versions": 100,
    "decisive_verdicts_with_registered_run_bound_evidence": 65
  },
  "privacy": "fixture data only; local PostgreSQL and encrypted local evidence; no hosted judge or Langfuse call"
}