{
  "bench_version": 1,
  "track": "vision",
  "model": "cnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "sweep": {
      "best_cfg": {
        "lr": 0.001
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001
          },
          "mean": 0.7862499803304672
        },
        {
          "cfg": {
            "lr": 0.003
          },
          "mean": 0.8006249815225601
        },
        {
          "cfg": {
            "lr": 0.006
          },
          "mean": 0.8237500041723251
        }
      ],
      "full": {
        "mean": 0.7849999815225601,
        "std": 0.03665719838879184,
        "per_seed": [
          0.7874999642372131,
          0.75,
          0.8349999785423279,
          0.7724999785423279,
          0.8274999856948853,
          0.8174999952316284,
          0.7249999642372131,
          0.7649999856948853
        ],
        "n": 8
      }
    },
    "full": {
      "mean": 0.7849999815225601,
      "std": 0.03665719838879184,
      "per_seed": [
        0.7874999642372131,
        0.75,
        0.8349999785423279,
        0.7724999785423279,
        0.8274999856948853,
        0.8174999952316284,
        0.7249999642372131,
        0.7649999856948853
      ],
      "n": 8
    }
  },
  "idea": {
    "lr": 0.001,
    "per_seed": [
      0.7899999618530273,
      0.7699999809265137,
      0.8399999737739563,
      0.762499988079071,
      0.824999988079071,
      0.8075000047683716,
      0.7475000023841858,
      0.7549999952316284
    ],
    "mean": 0.7871874868869781
  },
  "comparison": {
    "delta_mean": 0.00218750536441803,
    "idea_wins": 4,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.002499997615814209,
      0.019999980926513672,
      0.004999995231628418,
      -0.009999990463256836,
      -0.002499997615814209,
      -0.009999990463256836,
      0.022500038146972656,
      -0.009999990463256836
    ],
    "p_value": 0.64985,
    "mde": 0.01097274108925552,
    "mde_rel_pct": 1.3978014455456615,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "math_check": {
    "matrix_frobenius_residual": 0.0,
    "max_abs_energy_error": 1.4210854715202004e-14,
    "relative_energy_rmse": 1.4996245561724451e-16,
    "random_relative_rmse": 0.32222904703868205,
    "random_mean_relative_std": 0.3208364891796913
  },
  "idea_grid": [
    {
      "lr": 0.001,
      "mean": 0.7871874868869781
    },
    {
      "lr": 0.003,
      "mean": 0.7893749848008156
    },
    {
      "lr": 0.006,
      "mean": 0.8271874785423279
    }
  ],
  "mechanism_signature": {
    "prediction": "exact modeled energy has zero estimator variance; random signs have nonzero variance",
    "predicted_exact_relative_rmse": 0.0,
    "observed_mean_exact_relative_rmse": 5.16787475035585e-08,
    "observed_mean_random_relative_std": 0.3647565320134163,
    "per_seed": [
      {
        "seed": 0,
        "exact_rel_rmse": 0.0,
        "random_rel_std": 0.33480483293533325
      },
      {
        "seed": 1,
        "exact_rel_rmse": 1.1523898280074718e-07,
        "random_rel_std": 0.3534063696861267
      },
      {
        "seed": 2,
        "exact_rel_rmse": 0.0,
        "random_rel_std": 0.44275858998298645
      },
      {
        "seed": 3,
        "exact_rel_rmse": 0.0,
        "random_rel_std": 0.4704761505126953
      },
      {
        "seed": 4,
        "exact_rel_rmse": 0.0,
        "random_rel_std": 0.3139871060848236
      },
      {
        "seed": 5,
        "exact_rel_rmse": 1.5191113789114752e-07,
        "random_rel_std": 0.28184041380882263
      },
      {
        "seed": 6,
        "exact_rel_rmse": 1.4627985933657328e-07,
        "random_rel_std": 0.2919972240924835
      },
      {
        "seed": 7,
        "exact_rel_rmse": 0.0,
        "random_rel_std": 0.42878156900405884
      }
    ],
    "confirmed": true
  },
  "protocol_note": "Vision selected because the intervention is activation/architecture normalization; duplicate CNN layers and identical trainer settings, with only energy statistic changed."
}