{
  "@context": "https://schema.org",
  "@type": "Dataset",
  "name": "CivicProof Leaderboard",
  "version": "1.1.0",
  "published": "2026-07-28",
  "canonicalUrl": "https://marco-is-my-friend.marcohergee813.chatgpt.site/nation/leaderboard",
  "benchmark": {
    "name": "CivicProofBench 100",
    "version": "2.0.0",
    "caseCount": 100,
    "url": "/civic-proof-bench.json",
    "sha256": "8a8a9b1cdbd1596cc93cff186faa9d4181d8017a777a7462785a03f236cf71d6"
  },
  "metrics": {
    "exactReplay": "Percentage of cases matching expected article set, disposition, and score.",
    "articleSetAccuracy": "Percentage of cases with the exact expected constitutional article set.",
    "dispositionAccuracy": "Percentage of cases with the expected disposition.",
    "scoreMae": "Mean absolute error between produced and expected scores."
  },
  "verifiedRuns": [
    {
      "rank": 1,
      "runId": "CPR-REF-20260728-001",
      "system": "CIVIC-PROOF-1.0 reference engine",
      "provider": "Marco Hergi",
      "systemType": "deterministic lexical reference implementation",
      "evaluatedAt": "2026-07-28T18:00:00Z",
      "caseCount": 100,
      "exactReplay": 100,
      "articleSetAccuracy": 100,
      "dispositionAccuracy": 100,
      "scoreMae": 0,
      "evidence": {
        "benchmark": "/civic-proof-bench.json",
        "modelCard": "/civic-proof-model-card.json",
        "api": "/api/nation/evaluate",
        "report": "/nation/leaderboard/reference"
      },
      "boundary": "A reference engine reproducing its own published rules is a consistency baseline, not evidence of general intelligence, legal competence, ethical judgment, or superiority over an external model."
    },
    {
      "rank": 2,
      "runId": "github-models-mistral-ai-mistral-small-2503-30398743931",
      "system": "mistral-ai/mistral-small-2503",
      "provider": "GitHub Models",
      "systemType": "external open-weight language model",
      "verificationClass": "project-maintainer CI run; not independent third-party evaluation",
      "modelVersion": "1",
      "evaluatedAt": "2026-07-28T21:00:31.351Z",
      "caseCount": 100,
      "exactReplay": 34,
      "articleSetAccuracy": 99,
      "dispositionAccuracy": 55,
      "scoreMae": 14.92,
      "retries": 0,
      "exclusions": 0,
      "humanReview": "none",
      "evidence": {
        "workflowRun": "https://github.com/MARCCHERGGI/indie-metrics-site/actions/runs/30398743931",
        "directory": "https://github.com/MARCCHERGGI/indie-metrics-site/tree/main/civicproof/results/mistral-small-2503-2026-07-28",
        "run": "https://raw.githubusercontent.com/MARCCHERGGI/indie-metrics-site/main/civicproof/results/mistral-small-2503-2026-07-28/run.json",
        "report": "https://raw.githubusercontent.com/MARCCHERGGI/indie-metrics-site/main/civicproof/results/mistral-small-2503-2026-07-28/report.json",
        "rawResponses": "https://raw.githubusercontent.com/MARCCHERGGI/indie-metrics-site/main/civicproof/results/mistral-small-2503-2026-07-28/raw-responses.json",
        "modelCatalog": "https://github.com/marketplace/models/azureml-mistral/mistral-small-2503"
      },
      "normalization": "One finite score outside 0 through 100 was clamped to the nearest schema boundary; the raw value remains public.",
      "boundary": "Public open-book regression result executed by project-maintainer CI. It is not independent third-party validation and does not establish legal competence, ethical authority, safety, or general intelligence."
    }
  ],
  "awaitingIndependentRuns": [
    {
      "system": "OpenAI model",
      "status": "not-yet-submitted",
      "score": null
    },
    {
      "system": "Anthropic model",
      "status": "not-yet-submitted",
      "score": null
    },
    {
      "system": "Google model",
      "status": "not-yet-submitted",
      "score": null
    },
    {
      "system": "Open-source model",
      "status": "maintainer-run-published; independent-replication-not-yet-submitted",
      "score": null
    }
  ],
  "submissionRequirements": [
    "Run all 100 published cases without editing the inputs.",
    "Publish the exact model identifier, provider, date, system prompt, parameters, and raw outputs.",
    "Publish a stable HTTPS evidence URL and the benchmark SHA-256.",
    "Disclose human review, retries, tool use, and excluded cases.",
    "Consent to public verification before a score is listed."
  ],
  "boundary": "Missing independent runs are displayed as unverified, never as zero. The Mistral row is reproducible external-model evidence executed by project-maintainer CI, not independent third-party validation. No provider endorsement is implied."
}