{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.0015
        },
        "mean": 0.0007738938584225252
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.0004595852587954141
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 0.00018638806068338454
      }
    ],
    "full": {
      "mean": 0.00014496273979602847,
      "std": 5.6302020060222334e-05,
      "per_seed": [
        0.00022238190285861492,
        0.0002155364491045475,
        0.00011000921949744225,
        0.00019762467127293348,
        0.00014344952069222927,
        8.176809205906466e-05,
        6.902739551151171e-05,
        0.00011990466737188399
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.00014406874470296316,
    "std": 5.6322857127222856e-05,
    "per_seed": [
      0.00022284468286670744,
      0.00021447740436997265,
      0.00010853678395505995,
      0.00019537980551831424,
      0.00014281165204010904,
      8.075518417172134e-05,
      6.867131014587358e-05,
      0.00011907313455594704
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -8.93995093065314e-07,
    "idea_wins": 7,
    "n_pairs": 8,
    "per_seed_diffs": [
      4.627800080925226e-07,
      -1.0590447345748544e-06,
      -1.4724355423823e-06,
      -2.2448657546192408e-06,
      -6.378686521202326e-07,
      -1.0129078873433173e-06,
      -3.5608536563813686e-07,
      -8.315328159369528e-07
    ],
    "p_value": 0.0224,
    "mde": 6.629623320875621e-07,
    "mde_rel_pct": 0.4573329208735921,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "mechanism_signature": {
    "idea_lr_sweep": [
      {
        "cfg": {
          "lr": 0.0015
        },
        "mean": 0.0007737674750387669
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.0004588514202623628
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 0.00018530966917751357
      }
    ],
    "chosen_idea_cfg": {
      "lr": 0.006
    },
    "predicted_max_jacobian": 0.0,
    "observed_mean_abs_alpha_over_tau": 0.049265923735219985,
    "observed_residual_disturbance_corr_abs": 0.999999999999998,
    "n_trained_models": 8,
    "confirmed": true,
    "interpretation": "The matched bench has no latent actuator error; the simplicity prior should identify zero disturbance."
  }
}