{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.023774019442498684
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.01013653608970344
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 12
        },
        "mean": 0.003741882392205298
      }
    ],
    "full": {
      "mean": 0.003835578914731741,
      "std": 0.0008182271774829689,
      "per_seed": [
        0.004058382008224726,
        0.003425257047638297,
        0.00307759759016335,
        0.004406292922794819,
        0.004498522728681564,
        0.0026040049269795418,
        0.005265520885586739,
        0.003349053207784891
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.006128221459221095,
    "std": 0.0019430310597602655,
    "per_seed": [
      0.005886178929358721,
      0.003636509645730257,
      0.005781157873570919,
      0.010280938819050789,
      0.007521466817706823,
      0.0044521004892885685,
      0.006602167151868343,
      0.004865251947194338
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.002292642544489354,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.001827796921133995,
      0.00021125259809195995,
      0.002703560283407569,
      0.00587464589625597,
      0.003022944089025259,
      0.0018480955623090267,
      0.0013366462662816048,
      0.0015161987394094467
    ],
    "p_value": 0.0081,
    "mde": 0.0014061178480241382,
    "mde_rel_pct": 36.65985967916089,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "mechanism_signature": {
    "track_structure": "controlled pendulum multi-step dynamics; recurrent state updates",
    "scheduler": "largest high-bin hidden-state disagreement histogram",
    "clusters": 4,
    "histogram_bins": 4,
    "signature": {
      "prediction": "trained-model synchronous-vs-sequential discrepancy scales approximately eta^2",
      "rows": [
        {
          "eta": 0.2,
          "observed_linf": 0.023520752787590027,
          "eta2": 0.04000000000000001
        },
        {
          "eta": 0.4,
          "observed_linf": 0.09223496913909912,
          "eta2": 0.16000000000000003
        },
        {
          "eta": 0.6,
          "observed_linf": 0.20175543427467346,
          "eta2": 0.36
        },
        {
          "eta": 0.8,
          "observed_linf": 0.3461129665374756,
          "eta2": 0.6400000000000001
        }
      ],
      "eta2_ratio_range": [
        0.5408015102148055,
        0.5880188196897506
      ],
      "confirmed": true
    }
  },
  "runtime_sec": 430.6290466785431,
  "protocol_notes": {
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "dataset_sizes": [
      400,
      100
    ],
    "idea_grid": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.2794509418308735
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.012458630255423486
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 12
        },
        "mean": 0.006128221459221095
      }
    ],
    "selected_idea_cfg": {
      "lr": 0.01,
      "epochs": 12
    },
    "baseline_grid_union": [
      {
        "lr": 0.001,
        "epochs": 12
      },
      {
        "lr": 0.003,
        "epochs": 12
      },
      {
        "lr": 0.01,
        "epochs": 12
      }
    ]
  }
}