{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "epochs": 10,
      "branches": 1
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 10,
          "branches": 1
        },
        "mean": 0.0019458123424556106
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 10,
          "branches": 1
        },
        "mean": 0.000998065312160179
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 10,
          "branches": 1
        },
        "mean": 0.0007849872272345237
      }
    ],
    "full": {
      "mean": 0.0007333773064601701,
      "std": 0.00023119481143299165,
      "per_seed": [
        0.0007082645315676928,
        0.0011445400305092335,
        0.00047928086132742465,
        0.0008078634855337441,
        0.00035395112354308367,
        0.0006696216878481209,
        0.0007750760996714234,
        0.0009284206316806376
      ],
      "n": 8
    }
  },
  "idea": {
    "cfg": {
      "lr": 0.006,
      "epochs": 10,
      "branches": 5
    },
    "mean": 0.0008847052595228888,
    "std": 0.00026667546357193395,
    "per_seed": [
      0.0010930252028629184,
      0.0010985390981659293,
      0.0007570339366793633,
      0.001170001458376646,
      0.00035861978540197015,
      0.0010039987973868847,
      0.0006030820077285171,
      0.0009933417895808816
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.00015132795306271873,
    "idea_wins": 2,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0003847606712952256,
      -4.600093234330416e-05,
      0.0002777530753519386,
      0.00036213797284290195,
      4.66866185888648e-06,
      0.00033437710953876376,
      -0.00017199409194290638,
      6.4921157900244e-05
    ],
    "p_value": 0.08585,
    "mde": 0.00017902087041145347,
    "mde_rel_pct": 24.410473140427907,
    "verdict": "no significant win",
    "system_worked": false
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 10,
        "branches": 5
      },
      "mean": 0.00476446797256358,
      "std": 0.0010113119245195041,
      "per_seed": [
        0.004423358011990786,
        0.00587823148816824,
        0.006476645823568106,
        0.0045180427841842175,
        0.0030923711601644754,
        0.004275563172996044,
        0.005356233566999435,
        0.004095297772437334
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 10,
        "branches": 5
      },
      "mean": 0.0016875666624400765,
      "std": 0.00044316453275384946,
      "per_seed": [
        0.0008459890959784389,
        0.0019791238009929657,
        0.0021841360721737146,
        0.001931163715198636,
        0.0013305676402524114,
        0.0013080938952043653,
        0.0018005802994593978,
        0.002120878780260682
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 10,
        "branches": 5
      },
      "mean": 0.0008847052595228888,
      "std": 0.00026667546357193395,
      "per_seed": [
        0.0010930252028629184,
        0.0010985390981659293,
        0.0007570339366793633,
        0.001170001458376646,
        0.00035861978540197015,
        0.0010039987973868847,
        0.0006030820077285171,
        0.0009933417895808816
      ],
      "n": 8
    }
  ],
  "budget": {
    "n_train": 800,
    "n_test": 300,
    "epochs": 10,
    "batch": 128,
    "lr_union": [
      0.001,
      0.003,
      0.006
    ]
  },
  "mechanism_signature": {
    "trained_model": true,
    "soft_minus_hard_mean": 0.004501420073211193,
    "soft_minus_hard_std": 0.008464416489005089,
    "branch_spread_mean": 0.2787400186061859,
    "subset_refinement": [
      {
        "M": 2,
        "mean_abs_to_full": 0.5144225358963013
      },
      {
        "M": 3,
        "mean_abs_to_full": 0.035924121737480164
      },
      {
        "M": 5,
        "mean_abs_to_full": 0.0
      }
    ],
    "confirmed": true,
    "note": "values measured on the seed-0 trained dynamics model"
  }
}