{
  "bench_version": 1,
  "track": "sparse_cycle_compatibility",
  "model": "custom_gru",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.6574999913573265
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.5949999913573265
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.4999999776482582
      }
    ],
    "full": {
      "mean": 0.5137499831616879,
      "std": 0.1000234298210712,
      "per_seed": [
        0.5999999642372131,
        0.5399999618530273,
        0.5699999928474426,
        0.28999999165534973,
        0.4699999988079071,
        0.48499998450279236,
        0.6399999856948853,
        0.5149999856948853
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.55375000461936,
    "std": 0.09809911813863173,
    "per_seed": [
      0.6650000214576721,
      0.574999988079071,
      0.6299999952316284,
      0.33000001311302185,
      0.5199999809265137,
      0.5400000214576721,
      0.6349999904632568,
      0.5350000262260437
    ],
    "n": 8,
    "lr": 0.01,
    "lambda_orth": 0.0,
    "edge_violation_mean": 0.8190760239958763
  },
  "comparison": {
    "delta_mean": 0.04000002145767212,
    "idea_wins": 1,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.06500005722045898,
      0.0350000262260437,
      0.06000000238418579,
      0.04000002145767212,
      0.04999998211860657,
      0.05500003695487976,
      -0.004999995231628418,
      0.020000040531158447
    ],
    "p_value": 0.0169,
    "mde": 0.019481796815583544,
    "mde_rel_pct": 3.792077363329512,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "custom_track": {
    "name": "sparse_cycle_compatibility",
    "file": "bench/custom_tracks/sparse_cycle_compatibility.py",
    "domain": "masked_categorical_compatibility"
  },
  "protocol_notes": "8 paired seeds; registered graph-compatible track; identical GRU hidden width, epochs, batch, and lr union.",
  "mechanism_signature": {
    "predicted_effect": "orthogonality penalty lowers exclusivity-edge embedding inner products",
    "baseline_edge_violation": null,
    "idea_edge_violation": 0.8190760239958763,
    "observed_reduction": null,
    "confirmed": false
  },
  "idea_sweep": [
    {
      "lr": 0.001,
      "lambda_orth": 0.0,
      "mean": 0.7193749994039536,
      "std": 0.14231209834077435,
      "per_seed": [
        0.7799999713897705,
        0.8500000238418579,
        0.7400000095367432,
        0.36000001430511475,
        0.800000011920929,
        0.7799999713897705,
        0.6949999928474426,
        0.75
      ],
      "edge_violation_mean": 0.9312416091561317
    },
    {
      "lr": 0.001,
      "lambda_orth": 1.0,
      "mean": 0.7375000044703484,
      "std": 0.14866068540331184,
      "per_seed": [
        0.7799999713897705,
        0.8500000238418579,
        0.8550000190734863,
        0.36000001430511475,
        0.800000011920929,
        0.7799999713897705,
        0.7250000238418579,
        0.75
      ],
      "edge_violation_mean": 0.8652326688170433
    },
    {
      "lr": 0.001,
      "lambda_orth": 3.0,
      "mean": 0.7375000044703484,
      "std": 0.14866068540331184,
      "per_seed": [
        0.7799999713897705,
        0.8500000238418579,
        0.8550000190734863,
        0.36000001430511475,
        0.800000011920929,
        0.7799999713897705,
        0.7250000238418579,
        0.75
      ],
      "edge_violation_mean": 0.8657455369830132
    },
    {
      "lr": 0.003,
      "lambda_orth": 0.0,
      "mean": 0.6524999961256981,
      "std": 0.14166421585966793,
      "per_seed": [
        0.8050000071525574,
        0.7850000262260437,
        0.7749999761581421,
        0.36000001430511475,
        0.5199999809265137,
        0.6449999809265137,
        0.6899999976158142,
        0.6399999856948853
      ],
      "edge_violation_mean": 0.8980525210499763
    },
    {
      "lr": 0.003,
      "lambda_orth": 1.0,
      "mean": 0.7012499943375587,
      "std": 0.12996994825186,
      "per_seed": [
        0.8050000071525574,
        0.8650000095367432,
        0.824999988079071,
        0.42500001192092896,
        0.7049999833106995,
        0.6449999809265137,
        0.6899999976158142,
        0.6499999761581421
      ],
      "edge_violation_mean": 0.6635167449712753
    },
    {
      "lr": 0.003,
      "lambda_orth": 3.0,
      "mean": 0.7012499943375587,
      "std": 0.12996994825186,
      "per_seed": [
        0.8050000071525574,
        0.8650000095367432,
        0.824999988079071,
        0.42500001192092896,
        0.7049999833106995,
        0.6449999809265137,
        0.6899999976158142,
        0.6499999761581421
      ],
      "edge_violation_mean": 0.6590283587574959
    },
    {
      "lr": 0.01,
      "lambda_orth": 0.0,
      "mean": 0.55375000461936,
      "std": 0.09809911813863173,
      "per_seed": [
        0.6650000214576721,
        0.574999988079071,
        0.6299999952316284,
        0.33000001311302185,
        0.5199999809265137,
        0.5400000214576721,
        0.6349999904632568,
        0.5350000262260437
      ],
      "edge_violation_mean": 0.8190760239958763
    },
    {
      "lr": 0.01,
      "lambda_orth": 1.0,
      "mean": 0.5756249986588955,
      "std": 0.12212793455754742,
      "per_seed": [
        0.675000011920929,
        0.7699999809265137,
        0.6549999713897705,
        0.33000001311302185,
        0.5199999809265137,
        0.5400000214576721,
        0.5600000023841858,
        0.5550000071525574
      ],
      "edge_violation_mean": 0.22121139988303185
    },
    {
      "lr": 0.01,
      "lambda_orth": 3.0,
      "mean": 0.6031250022351742,
      "std": 0.11712539810482167,
      "per_seed": [
        0.675000011920929,
        0.7400000095367432,
        0.675000011920929,
        0.33000001311302185,
        0.6200000047683716,
        0.574999988079071,
        0.6549999713897705,
        0.5550000071525574
      ],
      "edge_violation_mean": 0.15788554586470127
    }
  ]
}