{
  "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": "mistral-nemo:12b",
    "quantization": "Q4_0",
    "parameter_size": "12.2B",
    "family": "llama",
    "digest_prefix": "e7e06d107c6c"
  },
  "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": 1791553762740,
  "finished_at_ms": 1791556361544,
  "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": 142.8,
    "mean_decisions": 199.1,
    "mean_hits": 0.0,
    "mean_bumps": 142.2,
    "known_wall_repeats": 1383,
    "outcomes": {
      "timeout": 10
    },
    "llm_calls": 1991,
    "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": 1299.0,
    "calibration": {
      "definition": "P(le déplacement annoncé ne heurte pas de mur) ; déplacements hors pass",
      "samples": 1991,
      "brier_score": 0.6426,
      "brier_reference_constant_half": 0.25,
      "mean_confidence": 0.9484,
      "observed_move_success_rate": 0.2858,
      "overconfidence": 0.6626,
      "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": 0.2664,
      "observed_rate_ci95_high": 0.306,
      "reliability_bins": [
        {
          "lower": 0.8,
          "upper": 0.9,
          "samples": 2,
          "mean_confidence": 0.85,
          "observed_rate": 0.5,
          "observed_rate_ci95_low": 0.0945,
          "observed_rate_ci95_high": 0.9055,
          "gap": 0.35,
          "sufficient": false
        },
        {
          "lower": 0.9,
          "upper": 1.0,
          "samples": 1989,
          "mean_confidence": 0.9485,
          "observed_rate": 0.2856,
          "observed_rate_ci95_low": 0.2661,
          "observed_rate_ci95_high": 0.3058,
          "gap": 0.6629,
          "sufficient": true
        }
      ],
      "ece": 0.6626,
      "ece_ci95_low": 0.6408,
      "ece_ci95_high": 0.6816,
      "mce": 0.6629,
      "murphy": {
        "reliability": 0.4391,
        "resolution": 0.0,
        "uncertainty": 0.2041,
        "within_bin_residual": -0.0005
      },
      "plain_language_fr": "Quand le modèle se dit sûr à environ 95 %, il a raison 29 % du temps (sur 1989 coups ; plage plausible 27 % à 31 %). En moyenne, il annonce 95 % et réussit 29 % de ses déplacements."
    }
  },
  "games": [
    {
      "seed": 0,
      "scenario_tag": "5327767cb2b8fa45",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 185,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 142,
      "unique_cells": 11,
      "revisits": 25,
      "immediate_reversals": 23,
      "llm_calls": 185,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 185,
      "brier_score": 0.7029,
      "duration_ms": 234394,
      "evidence_sha256": "a669b32b33a3237c1976e3bf293690a05fce2fe7deebf50051869f830b3e06b3"
    },
    {
      "seed": 1,
      "scenario_tag": "123f758b160ea0d4",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 163,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 145,
      "unique_cells": 4,
      "revisits": 10,
      "immediate_reversals": 10,
      "llm_calls": 163,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 163,
      "brier_score": 0.8313,
      "duration_ms": 222100,
      "evidence_sha256": "b24ad1352352490b989e8d10128ec9c23c1df5e7e4496d024ac6ebfa0d847391"
    },
    {
      "seed": 2,
      "scenario_tag": "1054437bd2bcae7b",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 78,
      "decisions": 300,
      "hits": 0,
      "bumps": 72,
      "known_wall_repeats": 68,
      "unique_cells": 5,
      "revisits": 224,
      "immediate_reversals": 194,
      "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": 0.2185,
      "duration_ms": 465222,
      "evidence_sha256": "2fadb381bd5e3786a9cea2c9223e3aa639b2fddc4f176bae0dd48eddf68533b0"
    },
    {
      "seed": 3,
      "scenario_tag": "992847cee040357d",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 154,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 146,
      "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": 0.8791,
      "duration_ms": 216192,
      "evidence_sha256": "76e2f01ad3112345994fcaf1bdb501843ca3e36dcf70e16bf3d26df8eed79f58"
    },
    {
      "seed": 4,
      "scenario_tag": "a3775211eaf136e7",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 273,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 147,
      "unique_cells": 5,
      "revisits": 119,
      "immediate_reversals": 116,
      "llm_calls": 273,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 273,
      "brier_score": 0.4974,
      "duration_ms": 318765,
      "evidence_sha256": "1feaba143c53132f3a93eff37086d60551283117a9dd747ccb20969af2f2be27"
    },
    {
      "seed": 5,
      "scenario_tag": "818402a1c2cdc4ee",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 177,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 146,
      "unique_cells": 6,
      "revisits": 22,
      "immediate_reversals": 18,
      "llm_calls": 177,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 177,
      "brier_score": 0.7652,
      "duration_ms": 231836,
      "evidence_sha256": "ab143e768f18dfb3ca82eeb5ce6c01a6679ca10f276b36e73990b55bf2bc288d"
    },
    {
      "seed": 6,
      "scenario_tag": "e3a35b35678367b5",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 152,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "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": 0.8913,
      "duration_ms": 181406,
      "evidence_sha256": "0a4a530fab26e36c008fb6e461935a17c2d8293336adcb92af1ea6bca77ef6fc"
    },
    {
      "seed": 7,
      "scenario_tag": "f2e49680b9a69523",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 152,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 146,
      "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": 0.8907,
      "duration_ms": 176787,
      "evidence_sha256": "f4f8029e0a4f3e86f183747c8d340be316862698f7c781896f7bcc69ab5bdda7"
    },
    {
      "seed": 8,
      "scenario_tag": "766646d7986ba8ef",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 283,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 147,
      "unique_cells": 5,
      "revisits": 129,
      "immediate_reversals": 128,
      "llm_calls": 283,
      "invalid_responses": 0,
      "timed_out_calls": 0,
      "timeout_retries": 0,
      "timeouts_after_retry": 0,
      "forced_passes": 0,
      "dropped_probability_fields": 0,
      "confidence_samples": 283,
      "brier_score": 0.4795,
      "duration_ms": 363266,
      "evidence_sha256": "92997544584e4605a850d147d3352914405501e2de98c7274039d240610a6db7"
    },
    {
      "seed": 9,
      "scenario_tag": "ac93e13596ea6c9a",
      "outcome": "timeout",
      "won": false,
      "picked_treasure": false,
      "rounds": 150,
      "decisions": 152,
      "hits": 0,
      "bumps": 150,
      "known_wall_repeats": 148,
      "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": 0.8896,
      "duration_ms": 178147,
      "evidence_sha256": "5d9839ca4d3743e1cac91107f18e1e28b6b3a2a0c3e68652f8c59325bf498963"
    }
  ],
  "report_sha256": "36df2ea08813aa25ee243269d5cfe5e2df8a4a399e7e6062a53434184f227154",
  "timeout_policy": "outilsia.decision_quality.timeout_policy.v2",
  "warmup": {
    "status": "ok",
    "duration_ms": 10507,
    "counted": false
  },
  "honesty_note": "Mesure sur votre machine, graines fixes ; ce n'est pas un classement universel des modèles.",
  "raw_model_output_stored": false
}