{
  "schema": "outilsia.decision_quality.report.v1",
  "status": "completed",
  "status_detail": null,
  "protocol_version": "dragon-uncertainty.v1",
  "bridge_schema": "outilsia.dragon.agent-bridge.v1",
  "prompt_version": "outilsia.decision_quality.prompt.v1",
  "prompt_sha256": "84637504c86b4f714d1b3756bcc63c5bcae572a3d4e2f3c3c995e41d6b144a4d",
  "scaffolding": "observation_jsonl_only.v1",
  "response_format": "ollama_format_json+bridge_decision.v1",
  "model": {
    "model": "hermes3:8b",
    "quantization": "Q4_0",
    "parameter_size": "8.0B",
    "family": "llama",
    "digest_prefix": "4f6b83f30b62"
  },
  "runtime": "ollama",
  "ollama_version": "0.18.2",
  "app_version": "0.1.2",
  "build_id": "local-dev",
  "config": {
    "seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7,
      8,
      9
    ],
    "move_timeout_seconds": 30,
    "max_rounds": 150,
    "max_decisions": 300,
    "max_attempts_per_decision": 3
  },
  "generation_options": {
    "num_ctx": 4096,
    "num_predict": 160,
    "seed": 42,
    "temperature": 0
  },
  "started_at_ms": 1791552453018,
  "finished_at_ms": 1791553726796,
  "summary": {
    "games": 10,
    "wins": 0,
    "win_rate": 0.0,
    "win_rate_ci95_low": 0.0,
    "win_rate_ci95_high": 0.2775,
    "treasure_pickups": 0,
    "treasure_rate": 0.0,
    "mean_rounds": 120.8,
    "mean_decisions": 125.8,
    "mean_hits": 0.6,
    "mean_bumps": 120.6,
    "known_wall_repeats": 1188,
    "outcomes": {
      "death_3hits": 2,
      "timeout": 8
    },
    "llm_calls": 1258,
    "invalid_responses": 0,
    "invalid_rate": 0.0,
    "invalid_by_reason": {},
    "timed_out_calls": 0,
    "timeout_retries": 0,
    "timeouts_after_retry": 0,
    "timeout_rate": 0.0,
    "timeout_warning_threshold": 0.05,
    "measurement_perturbed": false,
    "measurement_warning": null,
    "forced_passes": 0,
    "dropped_probability_fields": 0,
    "mean_latency_ms": 1004.0,
    "calibration": {
      "definition": "P(le déplacement annoncé ne heurte pas de mur) ; déplacements hors pass",
      "samples": 1258,
      "brier_score": 0.6152,
      "brier_reference_constant_half": 0.25,
      "mean_confidence": 0.7999,
      "observed_move_success_rate": 0.0413,
      "overconfidence": 0.7586,
      "method": "brier+reliability_10_bins+murphy_1973+ece(bootstrap_1000_mt19937_seed42)+wilson_per_bin.v2",
      "bins": 10,
      "min_samples_for_verdict": 30,
      "min_samples_per_bin": 5,
      "sample_sufficient": true,
      "observed_rate_ci95_low": 0.0317,
      "observed_rate_ci95_high": 0.0538,
      "reliability_bins": [
        {
          "lower": 0.7,
          "upper": 0.8,
          "samples": 1,
          "mean_confidence": 0.7,
          "observed_rate": 1.0,
          "observed_rate_ci95_low": 0.2065,
          "observed_rate_ci95_high": 1.0,
          "gap": -0.3,
          "sufficient": false
        },
        {
          "lower": 0.8,
          "upper": 0.9,
          "samples": 1257,
          "mean_confidence": 0.8,
          "observed_rate": 0.0406,
          "observed_rate_ci95_low": 0.031,
          "observed_rate_ci95_high": 0.053,
          "gap": 0.7594,
          "sufficient": true
        }
      ],
      "ece": 0.7591,
      "ece_ci95_low": 0.7479,
      "ece_ci95_high": 0.7698,
      "mce": 0.7594,
      "murphy": {
        "reliability": 0.5763,
        "resolution": 0.0007,
        "uncertainty": 0.0396,
        "within_bin_residual": 0.0
      },
      "plain_language_fr": "Quand le modèle se dit sûr à environ 80 %, il a raison 4 % du temps (sur 1257 coups ; plage plausible 3 % à 5 %). En moyenne, il annonce 80 % et réussit 4 % de ses déplacements."
    }
  },
  "games": [
    {
      "seed": 0,
      "scenario_tag": "5327767cb2b8fa45",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 4,
      "decisions": 15,
      "hits": 3,
      "bumps": 3,
      "known_wall_repeats": 2,
      "unique_cells": 5,
      "revisits": 8,
      "immediate_reversals": 0,
      "llm_calls": 15,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 15,
      "brier_score": 0.16,
      "duration_ms": 16903,
      "evidence_sha256": "62e8eb5ba91241587a85ca45e7ac244066563a4969454da23a2e647ced9fb815"
    },
    {
      "seed": 1,
      "scenario_tag": "123f758b160ea0d4",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 5,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 154,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 154,
      "brier_score": 0.6244,
      "duration_ms": 191655,
      "evidence_sha256": "fb416901283e1ea46c6f4d5f8e63bb5e129546cf73629eabe224be4ce503ba4f"
    },
    {
      "seed": 2,
      "scenario_tag": "1054437bd2bcae7b",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 4,
      "decisions": 15,
      "hits": 3,
      "bumps": 3,
      "known_wall_repeats": 2,
      "unique_cells": 5,
      "revisits": 8,
      "immediate_reversals": 0,
      "llm_calls": 15,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 15,
      "brier_score": 0.16,
      "duration_ms": 14137,
      "evidence_sha256": "a8e97c2a8ccb24af5317392bf69ac1849f227509c3b436f8cd44ef95409ea811"
    },
    {
      "seed": 3,
      "scenario_tag": "992847cee040357d",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 153,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 4,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 153,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 153,
      "brier_score": 0.6282,
      "duration_ms": 139610,
      "evidence_sha256": "f2f464d27171f5ff15908c28886719928ea47883ad47065b9f877a885fd75635"
    },
    {
      "seed": 4,
      "scenario_tag": "a3775211eaf136e7",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 5,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 154,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 154,
      "brier_score": 0.6244,
      "duration_ms": 144327,
      "evidence_sha256": "aa37d062ef3d122de14b055fb28df1d95f9c85c5b589ac913c9d67df2818072d"
    },
    {
      "seed": 5,
      "scenario_tag": "818402a1c2cdc4ee",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 155,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 6,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 155,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 155,
      "brier_score": 0.6206,
      "duration_ms": 160795,
      "evidence_sha256": "476b82a07059eb92b2fe87136ba10958748f7b4ace2ca5d51dae760e466a059b"
    },
    {
      "seed": 6,
      "scenario_tag": "e3a35b35678367b5",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 147,
      "unique_cells": 5,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 154,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 154,
      "brier_score": 0.6244,
      "duration_ms": 150585,
      "evidence_sha256": "0afaff6d2d9022479b27d3fa1ee133a9ff4f803d24444cb602cbc6c716d1ad67"
    },
    {
      "seed": 7,
      "scenario_tag": "f2e49680b9a69523",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 151,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "unique_cells": 2,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 151,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 151,
      "brier_score": 0.636,
      "duration_ms": 137853,
      "evidence_sha256": "2ec0efaf5f779a8a0a6d28d6c40101296faa09eda408b2a5d30af858275b1868"
    },
    {
      "seed": 8,
      "scenario_tag": "766646d7986ba8ef",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 5,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 154,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 154,
      "brier_score": 0.6247,
      "duration_ms": 151030,
      "evidence_sha256": "5b147aea026845ab426d6b20b3490af8995c190c59e0ee4d60d55c215fdab9e4"
    },
    {
      "seed": 9,
      "scenario_tag": "ac93e13596ea6c9a",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 153,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 4,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 153,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 153,
      "brier_score": 0.6282,
      "duration_ms": 157210,
      "evidence_sha256": "2c8ced24679900b48126b5c2f26fa714c83459bf81bbfbb9263ef38738f87458"
    }
  ],
  "report_sha256": "9a739479938cf89925f9479eab0ea59fa51736fbd77e3084e3f39f7fa7e6edc7",
  "timeout_policy": "outilsia.decision_quality.timeout_policy.v2",
  "warmup": {
    "status": "ok",
    "duration_ms": 9402,
    "counted": false
  },
  "honesty_note": "Mesure sur votre machine, graines fixes ; ce n'est pas un classement universel des modèles.",
  "raw_model_output_stored": false
}