{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 140.784441947937
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 870.165714263916
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 682.9689865112305
      }
    ],
    "full": {
      "mean": 118.6587815284729,
      "std": 53.34609536660658,
      "per_seed": [
        202.70230102539062,
        60.04500961303711,
        118.25131225585938,
        182.13914489746094,
        153.9148406982422,
        100.01561737060547,
        89.7911148071289,
        42.410911560058594
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.5461385995149612,
    "std": 0.43340968114270667,
    "per_seed": [
      0.30781564116477966,
      0.6033635139465332,
      0.3968035876750946,
      1.6646853685379028,
      0.3090420663356781,
      0.2753473222255707,
      0.3914429545402527,
      0.4206083416938782
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -118.11264292895794,
    "idea_wins": 8,
    "n_pairs": 8,
    "per_seed_diffs": [
      -202.39448538422585,
      -59.441646099090576,
      -117.85450866818428,
      -180.47445952892303,
      -153.6057986319065,
      -99.7402700483799,
      -89.39967185258865,
      -41.990303218364716
    ],
    "p_value": 0.0081,
    "mde": 47.5644061392987,
    "mde_rel_pct": 40.085028285820826,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "audit": {
    "structural_match": "dynamics stability/control",
    "base_architecture": "rnn_small GRU plus scalar Laurent readout",
    "idea_configs": [
      {
        "lr": 0.001
      },
      {
        "lr": 0.003
      },
      {
        "lr": 0.01
      }
    ],
    "baseline_grid": [
      {
        "lr": 0.001
      },
      {
        "lr": 0.003
      },
      {
        "lr": 0.01
      }
    ],
    "baseline_best_lr": 0.001,
    "idea_best_lr": 0.001,
    "epochs": 12,
    "n_train": 400,
    "n_test": 200,
    "parameter_parity": true
  },
  "mechanism_signature": {
    "prediction": "unsafe output scales as t^-2; safe output is bounded (t^0)",
    "observed": {
      "baseline": {
        "distances": [
          0.05,
          0.03,
          0.02,
          0.01
        ],
        "mean_abs_outputs": [
          1.7172616720199585,
          4.7458672523498535,
          10.651176452636719,
          42.497459411621094
        ],
        "loglog_slope": -1.99381250167288,
        "predicted_slope": -2.0,
        "absolute_slope_error": 0.006187498327119956
      },
      "idea": {
        "distances": [
          0.05,
          0.03,
          0.02,
          0.01
        ],
        "mean_abs_outputs": [
          1.0119659900665283,
          1.0068100690841675,
          1.0042619705200195,
          1.0017340183258057
        ],
        "loglog_slope": 0.00618747125927932,
        "predicted_slope": 0.0,
        "absolute_slope_error": 0.00618747125927932
      }
    },
    "confirmed": true
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001
      },
      "mean": 0.5461385995149612
    },
    {
      "cfg": {
        "lr": 0.003
      },
      "mean": 0.6001838333904743
    },
    {
      "cfg": {
        "lr": 0.01
      },
      "mean": 0.9318437203764915
    }
  ]
}