{
  "auditDate": "2026-10-24",
  "auditedAtUtc": "2026-10-23T18:50:13.511291+00:00",
  "analysisVersion": "Syncvals analysis v1.2 \u00b7 2026-10-24",
  "snapshotHash": "29b46d5ea53031efee082a7b9ff6f1c4535a37e21efce8924f9b05e6cdf2948f",
  "numericalReplay": {
    "status": "pass",
    "scalarOrStructureComparisons": 83006,
    "taskModelObjects": 1572,
    "modelDomainObjects": 238,
    "pairedNumericalObjects": 12,
    "bootstrapReplicates": 2000,
    "seed": 20261023
  },
  "mutationTests": {
    "unmodifiedFullReplay": "pass",
    "alteredModelCountDetected": true,
    "alteredRunRewardDetected": true,
    "alteredQualityGateDetected": true,
    "alteredMacroIntervalDetected": true,
    "alteredPairIntervalDetected": true
  },
  "sourceEvidence": {
    "sourceFilesHashMatched": 6,
    "trialRecordsMatchedToPublic": 13084,
    "snapshotValidRecords": 11300,
    "analyticValidRecords": 11299,
    "snapshotValidNamedAgentRecords": 8080,
    "analyticValidNamedAgentRecords": 8079,
    "eligibleNamedAgentLabels": 1643,
    "qualityExcludedTaskVersions": 371,
    "benchmarkTaskVersions": 2046,
    "reviewRecords": 286,
    "reviewedDistinctTrialIds": 276,
    "normalizedOutcomeReviewMatches": 249,
    "normalizedOutcomeReviewDisagreements": 0,
    "classifierModelFields": 0,
    "classifierPromptHashFields": 0,
    "reviewerIndependenceFields": 0,
    "timeoutBudgetFields": 0,
    "selectionFields": 0,
    "duplicateTrialIds": 0,
    "markedDuplicateCohortRows": 157,
    "validTimestampCollisionGroups": 185,
    "retiredEvidenceUrls": 1018,
    "remainingPublicEvidenceUrls": 0
  },
  "limitations": [
    "Numerical agreement does not establish that an original execution occurred or that its grade was correct.",
    "Validity, task-quality flags, model eras, and diagnostic-source labels remain declared inputs.",
    "Review-record consistency does not establish human authorship, independence, or correctness.",
    "Classifier prompt execution and reward-guard provenance cannot be attested from missing metadata.",
    "Bootstrap reproducibility does not establish population coverage or remove selection bias.",
    "Source fingerprints verify byte identity, not semantic truth."
  ],
  "sha256": {
    "verify-scores.py": "563c091e922e12b9a3e2e77ffeb024a23c7ab89a3354932807c8da9e09ac716c",
    "run-results.json": "39a4cab93b0624b1b74f0a9c6126de8c6f31c8be59d6fe7cc394b8ddcf19db3d",
    "task-catalog.json": "d6a93084a5925d8284acff51a0ec4bded11fadd986d83a481c14c36feffd7763",
    "model-scores.json": "e5b3511745c7d8c03a2c881ecf6c4bb918434df0e4c3a817dfb85d5b23a54812",
    "figure-data.json": "8e562f62f58406eb8fce67f74a385cb34e522444a210be7d90f412434c3a2632"
  }
}