{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.001864717691205442
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.0010901302302954718
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.0006186682076076977
      }
    ],
    "full": {
      "mean": 0.00054517226817552,
      "std": 0.00020361842727135583,
      "per_seed": [
        0.0004930912400595844,
        0.0004765945195686072,
        0.0004673648509196937,
        0.0010376222198829055,
        0.000511663849465549,
        0.00027233464061282575,
        0.0005571767687797546,
        0.0005455300561152399
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.07319906074553728,
    "std": 0.03402286097230311,
    "per_seed": [
      0.056350141763687134,
      0.08426478505134583,
      0.052617624402046204,
      0.06809338182210922,
      0.04265451803803444,
      0.04711783304810524,
      0.07909751683473587,
      0.15539668500423431
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.07265388847736176,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.05585705052362755,
      0.08378819053177722,
      0.05215025955112651,
      0.06705575960222632,
      0.04214285418856889,
      0.046845498407492414,
      0.07854034006595612,
      0.15485115494811907
    ],
    "p_value": 0.0081,
    "mde": 0.030388641902834822,
    "mde_rel_pct": 5574.135677248187,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "mechanism_signature": {
    "math_sanity": {
      "formula": "q95(|exp(.5s)-exp(.35s)| |x0|)",
      "rows": [
        {
          "epsilon": 0.01,
          "predicted": 0.055187317647128684,
          "observed": 0.056,
          "abs_error": 0.0008126823528713176
        },
        {
          "epsilon": 0.03,
          "predicted": 0.15844985968729142,
          "observed": 0.16,
          "abs_error": 0.0015501403127085867
        },
        {
          "epsilon": 0.08,
          "predicted": 0.3838849584642059,
          "observed": 0.384,
          "abs_error": 0.00011504153579411369
        }
      ],
      "passed": true
    },
    "trained_model_signature": {
      "prediction": "local q95 flow error should remain near epsilon before reset",
      "observed_mean_q95_by_window": [
        0.41497564166784273,
        0.4705420028418301,
        NaN,
        NaN,
        NaN,
        NaN,
        NaN,
        NaN,
        NaN,
        NaN
      ],
      "epsilon": 0.1,
      "confirmed": false
    },
    "idea_config_grid": [
      {
        "lr": 0.001
      },
      {
        "lr": 0.003
      },
      {
        "lr": 0.01
      }
    ],
    "idea_meta": {
      "0": {
        "windows": 2,
        "q95_errors": [
          0.20428025126457203,
          0.43705364465713503
        ],
        "epsilon": 0.1
      },
      "1": {
        "windows": 2,
        "q95_errors": [
          0.5327116042375563,
          0.5056339323520659
        ],
        "epsilon": 0.1
      },
      "2": {
        "windows": 2,
        "q95_errors": [
          0.28441762030124623,
          0.4018018126487727
        ],
        "epsilon": 0.1
      },
      "3": {
        "windows": 2,
        "q95_errors": [
          0.5256956934928891,
          0.401323390007019
        ],
        "epsilon": 0.1
      },
      "4": {
        "windows": 2,
        "q95_errors": [
          0.1993324041366577,
          0.3880627930164337
        ],
        "epsilon": 0.1
      },
      "5": {
        "windows": 2,
        "q95_errors": [
          0.4679859936237335,
          0.3935546219348907
        ],
        "epsilon": 0.1
      },
      "6": {
        "windows": 2,
        "q95_errors": [
          0.4792444229125976,
          0.5255674421787261
        ],
        "epsilon": 0.1
      },
      "7": {
        "windows": 2,
        "q95_errors": [
          0.6261371433734889,
          0.7113383859395978
        ],
        "epsilon": 0.1
      }
    }
  }
}