{
  "bench_version": 1,
  "track": "sequence",
  "model": "transformer_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001,
      "epochs": 2,
      "K": 4,
      "tau": 0.7,
      "alpha": 0.5
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 2,
          "K": 4,
          "tau": 0.7,
          "alpha": 0.5
        },
        "mean": 0.9283310621976852
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 2,
          "K": 4,
          "tau": 0.7,
          "alpha": 0.5
        },
        "mean": 1.0033762753009796
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 2,
          "K": 4,
          "tau": 0.7,
          "alpha": 0.5
        },
        "mean": 1.0917343497276306
      }
    ],
    "full": {
      "mean": 0.9580474123358727,
      "std": 0.07569227656456869,
      "per_seed": [
        0.9577249884605408,
        0.8674030900001526,
        0.9789561033248901,
        0.9092400670051575,
        1.1384438276290894,
        0.9376952052116394,
        0.9597494006156921,
        0.9151666164398193
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.9508912265300751,
    "std": 0.05702490985176115,
    "per_seed": [
      0.9412082433700562,
      0.8438071012496948,
      0.9839955568313599,
      0.9183895587921143,
      1.0529438257217407,
      0.9422717094421387,
      0.9901344180107117,
      0.9343793988227844
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.007156185805797577,
    "idea_wins": 3,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.01651674509048462,
      -0.023595988750457764,
      0.0050394535064697266,
      0.009149491786956787,
      -0.08550000190734863,
      0.004576504230499268,
      0.03038501739501953,
      0.019212782382965088
    ],
    "p_value": 0.70955,
    "mde": 0.030234908489597815,
    "mde_rel_pct": 3.1558885395744953,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "connected bipartite tree retains 2L-1 edges independent of K",
    "predicted_tree_edges": 63.0,
    "observed_tree_edges": 63.0,
    "edge_count_error": 0.0,
    "confirmed": true,
    "measured_on_trained_model": true
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 2,
        "K": 4,
        "tau": 0.7,
        "alpha": 0.5
      },
      "mean": 0.9508912265300751,
      "std": 0.05702490985176115,
      "per_seed": [
        0.9412082433700562,
        0.8438071012496948,
        0.9839955568313599,
        0.9183895587921143,
        1.0529438257217407,
        0.9422717094421387,
        0.9901344180107117,
        0.9343793988227844
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 2,
        "K": 4,
        "tau": 0.7,
        "alpha": 0.5
      },
      "mean": 0.9828187823295593,
      "std": 0.03844168507903302,
      "per_seed": [
        0.9239319562911987,
        1.0003299713134766,
        1.0384844541549683,
        0.9352728724479675,
        0.9620147347450256,
        0.9814748167991638,
        1.0294827222824097,
        0.9915587306022644
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 2,
        "K": 4,
        "tau": 0.7,
        "alpha": 0.5
      },
      "mean": 1.1042008697986603,
      "std": 0.12163079384084556,
      "per_seed": [
        1.053992748260498,
        0.8888604640960693,
        1.2662471532821655,
        1.2502007484436035,
        1.0634373426437378,
        1.0349705219268799,
        1.2190437316894531,
        1.056854248046875
      ],
      "n": 8
    }
  ]
}