{
  "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": "qwen2.5:14b",
    "quantization": "Q4_K_M",
    "parameter_size": "14.8B",
    "family": "qwen2",
    "digest_prefix": "7cdf5a0187d5"
  },
  "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": 1791556397537,
  "finished_at_ms": 1791558374306,
  "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": 96.7,
    "mean_decisions": 149.7,
    "mean_hits": 0.9,
    "mean_bumps": 91.1,
    "known_wall_repeats": 884,
    "outcomes": {
      "death_3hits": 3,
      "timeout": 7
    },
    "llm_calls": 1497,
    "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": 1311.0,
    "calibration": {
      "definition": "P(le déplacement annoncé ne heurte pas de mur) ; déplacements hors pass",
      "samples": 1497,
      "brier_score": 0.352,
      "brier_reference_constant_half": 0.25,
      "mean_confidence": 0.748,
      "observed_move_success_rate": 0.3914,
      "overconfidence": 0.3565,
      "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": 0.367,
      "observed_rate_ci95_high": 0.4164,
      "reliability_bins": [
        {
          "lower": 0.5,
          "upper": 0.6,
          "samples": 165,
          "mean_confidence": 0.5,
          "observed_rate": 0.0121,
          "observed_rate_ci95_low": 0.0033,
          "observed_rate_ci95_high": 0.0431,
          "gap": 0.4879,
          "sufficient": true
        },
        {
          "lower": 0.7,
          "upper": 0.8,
          "samples": 1032,
          "mean_confidence": 0.75,
          "observed_rate": 0.4496,
          "observed_rate_ci95_low": 0.4195,
          "observed_rate_ci95_high": 0.4801,
          "gap": 0.3004,
          "sufficient": true
        },
        {
          "lower": 0.8,
          "upper": 0.9,
          "samples": 223,
          "mean_confidence": 0.85,
          "observed_rate": 0.3229,
          "observed_rate_ci95_low": 0.2649,
          "observed_rate_ci95_high": 0.3868,
          "gap": 0.5271,
          "sufficient": true
        },
        {
          "lower": 0.9,
          "upper": 1.0,
          "samples": 77,
          "mean_confidence": 0.9571,
          "observed_rate": 0.6234,
          "observed_rate_ci95_low": 0.5117,
          "observed_rate_ci95_high": 0.7233,
          "gap": 0.3338,
          "sufficient": true
        }
      ],
      "ece": 0.3565,
      "ece_ci95_low": 0.3329,
      "ece_ci95_high": 0.3801,
      "mce": 0.5271,
      "murphy": {
        "reliability": 0.1356,
        "resolution": 0.0217,
        "uncertainty": 0.2382,
        "within_bin_residual": -0.0001
      },
      "plain_language_fr": "Quand le modèle se dit sûr à environ 75 %, il a raison 45 % du temps (sur 1032 coups ; plage plausible 42 % à 48 %). En moyenne, il annonce 75 % et réussit 39 % de ses déplacements."
    }
  },
  "games": [
    {
      "seed": 0,
      "scenario_tag": "5327767cb2b8fa45",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 267,
      "hits": 0,
      "bumps": 140,
      "known_wall_repeats": 138,
      "unique_cells": 4,
      "revisits": 124,
      "immediate_reversals": 124,
      "llm_calls": 267,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 267,
      "brier_score": 0.3218,
      "duration_ms": 353932,
      "evidence_sha256": "cf514245fbfafcbbc5354394f803e4ba7a0e2f3a946eeb3a4e9417de833e4a08"
    },
    {
      "seed": 1,
      "scenario_tag": "123f758b160ea0d4",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 162,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 145,
      "unique_cells": 11,
      "revisits": 2,
      "immediate_reversals": 1,
      "llm_calls": 162,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 162,
      "brier_score": 0.6491,
      "duration_ms": 189943,
      "evidence_sha256": "afacd82dca6288d9c62bfe47d159b27274ae5cea8a617a1c220899f497b7edd1"
    },
    {
      "seed": 2,
      "scenario_tag": "1054437bd2bcae7b",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 4,
      "decisions": 24,
      "hits": 3,
      "bumps": 0,
      "known_wall_repeats": 0,
      "unique_cells": 6,
      "revisits": 19,
      "immediate_reversals": 11,
      "llm_calls": 24,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 24,
      "brier_score": 0.0059,
      "duration_ms": 41254,
      "evidence_sha256": "943af64fc95c28c284a71befc97a31e2e62b1676005a07f5967aca1615d8cd1b"
    },
    {
      "seed": 3,
      "scenario_tag": "992847cee040357d",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 155,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 146,
      "unique_cells": 4,
      "revisits": 2,
      "immediate_reversals": 2,
      "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": 0.5496,
      "duration_ms": 199013,
      "evidence_sha256": "d237e64097b8bdcc3a1a96bb6ef1aeb23fdc321863cd806c49222e7f341ed096"
    },
    {
      "seed": 4,
      "scenario_tag": "a3775211eaf136e7",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 175,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "unique_cells": 6,
      "revisits": 20,
      "immediate_reversals": 20,
      "llm_calls": 175,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 175,
      "brier_score": 0.4932,
      "duration_ms": 196163,
      "evidence_sha256": "e806a318a10a4f80e686a8094e65e3bd9af374266bc4c85af3b77abf597932a7"
    },
    {
      "seed": 5,
      "scenario_tag": "818402a1c2cdc4ee",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 216,
      "hits": 0,
      "bumps": 143,
      "known_wall_repeats": 139,
      "unique_cells": 8,
      "revisits": 66,
      "immediate_reversals": 64,
      "llm_calls": 216,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 216,
      "brier_score": 0.4124,
      "duration_ms": 280702,
      "evidence_sha256": "29ca3897047221fcd9a2028bff9bf32ddc4060d1e43b6e174579887a49d46339"
    },
    {
      "seed": 6,
      "scenario_tag": "e3a35b35678367b5",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 46,
      "decisions": 300,
      "hits": 0,
      "bumps": 12,
      "known_wall_repeats": 11,
      "unique_cells": 7,
      "revisits": 282,
      "immediate_reversals": 282,
      "llm_calls": 300,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 300,
      "brier_score": 0.0645,
      "duration_ms": 436213,
      "evidence_sha256": "886a593187455bbf4d83b9b74f3647363a44e81ad68fe86f09cac1d397719db6"
    },
    {
      "seed": 7,
      "scenario_tag": "f2e49680b9a69523",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 8,
      "decisions": 26,
      "hits": 3,
      "bumps": 7,
      "known_wall_repeats": 2,
      "unique_cells": 10,
      "revisits": 10,
      "immediate_reversals": 5,
      "llm_calls": 26,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 26,
      "brier_score": 0.2264,
      "duration_ms": 41833,
      "evidence_sha256": "f05aee58f87243fae432984737c7f050e7719632d4df14c25a69927bc3d91d0d"
    },
    {
      "seed": 8,
      "scenario_tag": "766646d7986ba8ef",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 9,
      "decisions": 20,
      "hits": 3,
      "bumps": 9,
      "known_wall_repeats": 6,
      "unique_cells": 10,
      "revisits": 2,
      "immediate_reversals": 0,
      "llm_calls": 20,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 20,
      "brier_score": 0.3303,
      "duration_ms": 31655,
      "evidence_sha256": "5f149e55f3453b00ff9ca6af69cb5a3a4cadfb0f98e9a1512b8323c94bedf38d"
    },
    {
      "seed": 9,
      "scenario_tag": "ac93e13596ea6c9a",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 152,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "unique_cells": 3,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 152,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 152,
      "brier_score": 0.2853,
      "duration_ms": 192932,
      "evidence_sha256": "9b78e1fc6feb04294a33d9276be68469a5da91f50513475321887770ec8f004d"
    }
  ],
  "report_sha256": "17819f935c0631a60f8cdc373557c7bffc1ad061f5597a20488bfee842a4af12",
  "timeout_policy": "outilsia.decision_quality.timeout_policy.v2",
  "warmup": {
    "status": "ok",
    "duration_ms": 12961,
    "counted": false
  },
  "honesty_note": "Mesure sur votre machine, graines fixes ; ce n'est pas un classement universel des modèles.",
  "raw_model_output_stored": false
}