{
  "bench_version": 1,
  "track": "vision",
  "model": "cnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.88
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.88625
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.895
      }
    ],
    "full": {
      "mean": 0.88375,
      "std": 0.025829972899714785,
      "per_seed": [
        0.9075,
        0.885,
        0.9075,
        0.82,
        0.8875,
        0.89,
        0.8925,
        0.88
      ],
      "n": 8
    },
    "idea_grid": [
      {
        "lr": 0.001,
        "result": {
          "mean": 0.8799999999999999,
          "std": 0.0347311099736245,
          "per_seed": [
            0.9075,
            0.8825,
            0.9075,
            0.8225
          ],
          "n": 4
        }
      },
      {
        "lr": 0.003,
        "result": {
          "mean": 0.881875,
          "std": 0.04258722666481113,
          "per_seed": [
            0.9,
            0.9275,
            0.8875,
            0.8125
          ],
          "n": 4
        }
      },
      {
        "lr": 0.01,
        "result": {
          "mean": 0.895,
          "std": 0.012119199643540817,
          "per_seed": [
            0.9,
            0.9075,
            0.875,
            0.8975
          ],
          "n": 4
        }
      }
    ]
  },
  "idea": {
    "mean": 0.8843749999999999,
    "std": 0.02533371024938904,
    "per_seed": [
      0.9075,
      0.8825,
      0.9075,
      0.8225,
      0.8875,
      0.89,
      0.8975,
      0.88
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.0006250000000000006,
    "idea_wins": 1,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0,
      -0.0025000000000000577,
      0.0,
      0.0025000000000000577,
      0.0,
      0.0,
      0.0050000000000000044,
      0.0
    ],
    "p_value": 0.7525,
    "mde": 0.0018529277478734423,
    "mde_rel_pct": 0.20966650612429333,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "fused mean e-value equals uncertainty-weighted mean local e-values",
    "observed": [
      {
        "accuracy": 0.0925,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 0.9467349927047926,
        "predicted_mean_true_e": 0.9467348347498856,
        "convexity_abs_error": 1.5795490693193415e-07
      },
      {
        "accuracy": 0.1175,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 1.0163047070950022,
        "predicted_mean_true_e": 1.0163049215843074,
        "convexity_abs_error": 2.1448930520584497e-07
      },
      {
        "accuracy": 0.0925,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 1.047165460267897,
        "predicted_mean_true_e": 1.0471653046459115,
        "convexity_abs_error": 1.556219855203267e-07
      },
      {
        "accuracy": 0.1775,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 0.9733103156789212,
        "predicted_mean_true_e": 0.973310362700516,
        "convexity_abs_error": 4.702159472902423e-08
      },
      {
        "accuracy": 0.1125,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 0.9805528636845341,
        "predicted_mean_true_e": 0.9805527689498554,
        "convexity_abs_error": 9.473467865728225e-08
      },
      {
        "accuracy": 0.11,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 0.976847474581956,
        "predicted_mean_true_e": 0.9768475761609023,
        "convexity_abs_error": 1.0157894636542153e-07
      },
      {
        "accuracy": 0.1025,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 1.0040690733188393,
        "predicted_mean_true_e": 1.0040692552907968,
        "convexity_abs_error": 1.819719575024692e-07
      },
      {
        "accuracy": 0.12,
        "coverage": 1.0,
        "avg_set_size": 10.0,
        "mean_true_e": 0.9969966644785999,
        "predicted_mean_true_e": 0.9969968797114046,
        "convexity_abs_error": 2.1523280469004646e-07
      }
    ],
    "confirmed": true,
    "max_abs_error": 2.1523280469004646e-07
  },
  "idea_sweep": [
    {
      "lr": 0.001,
      "result": {
        "mean": 0.8799999999999999,
        "std": 0.0347311099736245,
        "per_seed": [
          0.9075,
          0.8825,
          0.9075,
          0.8225
        ],
        "n": 4
      }
    },
    {
      "lr": 0.003,
      "result": {
        "mean": 0.881875,
        "std": 0.04258722666481113,
        "per_seed": [
          0.9,
          0.9275,
          0.8875,
          0.8125
        ],
        "n": 4
      }
    },
    {
      "lr": 0.01,
      "result": {
        "mean": 0.895,
        "std": 0.012119199643540817,
        "per_seed": [
          0.9,
          0.9075,
          0.875,
          0.8975
        ],
        "n": 4
      }
    }
  ],
  "notes": "Vision selected because labels/candidate-label e-values are classification-specific; N_TRAIN/N_TEST=400 and 3 epochs were used for budget."
}