{
  "bench_version": 1,
  "track": "sequence",
  "model": "custom_recurrent_state",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 1.100467175245285
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 1.0823257267475128
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 1.043792873620987
      }
    ],
    "full": {
      "mean": 1.0421477407217026,
      "std": 0.10101764442672699,
      "per_seed": [
        1.1993988752365112,
        0.8459368944168091,
        1.0848031044006348,
        1.0450326204299927,
        1.0629944801330566,
        1.080702543258667,
        1.0881637334823608,
        0.9301496744155884
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 1.0151974856853485,
    "std": 0.08790083123237098,
    "per_seed": [
      1.1760079860687256,
      0.8348132967948914,
      1.027022123336792,
      0.9901270866394043,
      1.0220698118209839,
      0.9881404042243958,
      1.0568915605545044,
      1.0265076160430908
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.026950255036354065,
    "idea_wins": 7,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.023390889167785645,
      -0.011123597621917725,
      -0.05778098106384277,
      -0.05490553379058838,
      -0.040924668312072754,
      -0.09256213903427124,
      -0.031272172927856445,
      0.09635794162750244
    ],
    "p_value": 0.21995,
    "mde": 0.04657542404598468,
    "mde_rel_pct": 4.46917670365341,
    "verdict": "no significant win",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "reversal output difference is explained by trained area-head contribution",
    "n": 64,
    "observed_output_reversal_rms": 0.28506407141685486,
    "predicted_area_head_rms": 0.28506410121917725,
    "observed_to_predicted_rms_ratio": 0.9999998954535234,
    "correlation": 0.9999999999999685,
    "confirmed": true,
    "model_test_mse": 1.1760079860687256,
    "idea_sweep": {
      "best_cfg": {
        "lr": 0.01
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001
          },
          "mean": 1.0460671931505203
        },
        {
          "cfg": {
            "lr": 0.003
          },
          "mean": 1.0292403250932693
        },
        {
          "cfg": {
            "lr": 0.01
          },
          "mean": 1.0069926232099533
        }
      ],
      "full": {
        "mean": 1.0151974856853485,
        "std": 0.08790083123237098,
        "per_seed": [
          1.1760079860687256,
          0.8348132967948914,
          1.027022123336792,
          0.9901270866394043,
          1.0220698118209839,
          0.9881404042243958,
          1.0568915605545044,
          1.0265076160430908
        ],
        "n": 8
      }
    }
  },
  "protocol_notes": {
    "architecture": "matched projection/state/readout; only area channels replace additive state",
    "epochs": 15,
    "n_train": 400,
    "n_test": 400,
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "baseline_and_idea_lr_union": [
      {
        "lr": 0.001
      },
      {
        "lr": 0.003
      },
      {
        "lr": 0.01
      }
    ],
    "baseline_selection_seeds": [
      0,
      1,
      2,
      3
    ],
    "parameters_baseline": 65,
    "parameters_idea": 185
  }
}