{
  "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:latest",
    "quantization": "Q4_K_M",
    "parameter_size": "7.6B",
    "family": "qwen2",
    "digest_prefix": "845dbda0ea48"
  },
  "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": 1791549588914,
  "finished_at_ms": 1791551335057,
  "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": 135.4,
    "mean_decisions": 139.5,
    "mean_hits": 0.3,
    "mean_bumps": 135.3,
    "known_wall_repeats": 1343,
    "outcomes": {
      "death_3hits": 1,
      "timeout": 9
    },
    "llm_calls": 1395,
    "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": 1244.0,
    "calibration": {
      "definition": "P(le déplacement annoncé ne heurte pas de mur) ; déplacements hors pass",
      "samples": 1395,
      "brier_score": 0.8804,
      "brier_reference_constant_half": 0.25,
      "mean_confidence": 0.9531,
      "observed_move_success_rate": 0.0301,
      "overconfidence": 0.923,
      "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": 0.0224,
      "observed_rate_ci95_high": 0.0404,
      "reliability_bins": [
        {
          "lower": 0.9,
          "upper": 1.0,
          "samples": 1395,
          "mean_confidence": 0.9531,
          "observed_rate": 0.0301,
          "observed_rate_ci95_low": 0.0224,
          "observed_rate_ci95_high": 0.0404,
          "gap": 0.923,
          "sufficient": true
        }
      ],
      "ece": 0.923,
      "ece_ci95_low": 0.9137,
      "ece_ci95_high": 0.9314,
      "mce": 0.923,
      "murphy": {
        "reliability": 0.8519,
        "resolution": 0.0,
        "uncertainty": 0.0292,
        "within_bin_residual": -0.0008
      },
      "plain_language_fr": "Quand le modèle se dit sûr à environ 95 %, il a raison 3 % du temps (sur 1395 coups ; plage plausible 2 % à 4 %). En moyenne, il annonce 95 % et réussit 3 % de ses déplacements."
    }
  },
  "games": [
    {
      "seed": 0,
      "scenario_tag": "5327767cb2b8fa45",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 155,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "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": 0.812,
      "duration_ms": 199149,
      "evidence_sha256": "92ebf03c9abf3493dee5c6a3e6162f0451bbeb21061ed9939d0729e6e6fbadcb"
    },
    {
      "seed": 1,
      "scenario_tag": "123f758b160ea0d4",
      "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": 0.8898,
      "duration_ms": 265161,
      "evidence_sha256": "a9ec305d626735d8db013d02f80a1acd8fbcf1eb44e5728a71925df9aab24f31"
    },
    {
      "seed": 2,
      "scenario_tag": "1054437bd2bcae7b",
      "outcome": "death_3hits",
      "won": false,
      "picked_treasure": false,
      "rounds": 4,
      "decisions": 12,
      "hits": 3,
      "bumps": 3,
      "known_wall_repeats": 2,
      "unique_cells": 4,
      "revisits": 6,
      "immediate_reversals": 0,
      "llm_calls": 12,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 12,
      "brier_score": 0.2381,
      "duration_ms": 16709,
      "evidence_sha256": "1192909c26654160601dffca7c7140c9901e0aee11d65c185cb586b50463c842"
    },
    {
      "seed": 3,
      "scenario_tag": "992847cee040357d",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "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": 0.9126,
      "duration_ms": 194597,
      "evidence_sha256": "bc412bf5dcd40f208107eee2d1b7305f829e82de6179fa356e116ac908af581b"
    },
    {
      "seed": 4,
      "scenario_tag": "a3775211eaf136e7",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 155,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "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": 0.7937,
      "duration_ms": 195370,
      "evidence_sha256": "5dcdd76a0fd8d8fe6311573a1e2f538bf8c98af24b9849fa551ed2245338585a"
    },
    {
      "seed": 5,
      "scenario_tag": "818402a1c2cdc4ee",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "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": 0.9139,
      "duration_ms": 167618,
      "evidence_sha256": "610a570574a7efca57cbf511225667337dbf3477a2ecaccc0ce60c69ac24103a"
    },
    {
      "seed": 6,
      "scenario_tag": "e3a35b35678367b5",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "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": 0.8911,
      "duration_ms": 163478,
      "evidence_sha256": "10f22d52d24cf3de3789b8b6cda54227ad1ad18351a2286178683550034bf885"
    },
    {
      "seed": 7,
      "scenario_tag": "f2e49680b9a69523",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 157,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 149,
      "unique_cells": 8,
      "revisits": 0,
      "immediate_reversals": 0,
      "llm_calls": 157,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 157,
      "brier_score": 0.9554,
      "duration_ms": 141391,
      "evidence_sha256": "26b83ff825f77441f533d529726856916e0ca32ba19dfa75f64fd70a3355d497"
    },
    {
      "seed": 8,
      "scenario_tag": "766646d7986ba8ef",
      "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": 0.9125,
      "duration_ms": 180509,
      "evidence_sha256": "a83c0493ebceb55ae79303505d3d46a432d516181f4830d5dbaca5ed8037cc90"
    },
    {
      "seed": 9,
      "scenario_tag": "ac93e13596ea6c9a",
      "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": 0.8931,
      "duration_ms": 211904,
      "evidence_sha256": "a537ae3445c3c55fd05454242356ac242c53bea1a2e3f4278df11cff70d72895"
    }
  ],
  "report_sha256": "c837922ac1b5ff0d51f5ed2d9c535315a969875ae71c4044a474a979b79227c1",
  "timeout_policy": "outilsia.decision_quality.timeout_policy.v2",
  "warmup": {
    "status": "ok",
    "duration_ms": 10026,
    "counted": false
  },
  "honesty_note": "Mesure sur votre machine, graines fixes ; ce n'est pas un classement universel des modèles.",
  "raw_model_output_stored": false
}