1{
  2  "bench_version": 1,
  3  "track": "rooted_tree_branch_classification",
  4  "model": "mlp_tiny",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.003
 10    },
 11    "sweep": [
 12      {
 13        "cfg": {
 14          "lr": 0.001
 15        },
 16        "mean": 0.5112499818205833
 17      },
 18      {
 19        "cfg": {
 20          "lr": 0.003
 21        },
 22        "mean": 0.4912499859929085
 23      },
 24      {
 25        "cfg": {
 26          "lr": 0.01
 27        },
 28        "mean": 0.4912499859929085
 29      }
 30    ],
 31    "full": {
 32      "mean": 0.47624998912215233,
 33      "std": 0.04292071696077957,
 34      "per_seed": [
 35        0.4599999785423279,
 36        0.5450000166893005,
 37        0.5299999713897705,
 38        0.429999977350235,
 39        0.429999977350235,
 40        0.4399999976158142,
 41        0.4650000035762787,
 42        0.5099999904632568
 43      ],
 44      "n": 8
 45    }
 46  },
 47  "idea": {
 48    "mean": 0.0,
 49    "std": 0.0,
 50    "per_seed": [
 51      0.0,
 52      0.0,
 53      0.0,
 54      0.0,
 55      0.0,
 56      0.0,
 57      0.0,
 58      0.0
 59    ],
 60    "n": 8
 61  },
 62  "comparison": {
 63    "delta_mean": -0.47624998912215233,
 64    "idea_wins": 8,
 65    "n_pairs": 8,
 66    "per_seed_diffs": [
 67      -0.4599999785423279,
 68      -0.5450000166893005,
 69      -0.5299999713897705,
 70      -0.429999977350235,
 71      -0.429999977350235,
 72      -0.4399999976158142,
 73      -0.4650000035762787,
 74      -0.5099999904632568
 75    ],
 76    "p_value": 0.0081,
 77    "mde": 0.038366227085568125,
 78    "mde_rel_pct": 8.05590088438357,
 79    "verdict": "idea better (significant)",
 80    "system_worked": true
 81  },
 82  "custom_track": {
 83    "name": "rooted_tree_branch_classification",
 84    "file": "tree_polynomial_track.py",
 85    "domain": "graph-nn"
 86  },
 87  "idea_config": {
 88    "lr": 0.01
 89  },
 90  "idea_sweep": [
 91    {
 92      "cfg": {
 93        "lr": 0.001
 94      },
 95      "result": {
 96        "mean": 0.026249999413266778,
 97        "std": 0.022185299058062068,
 98        "per_seed": [
 99          0.004999999888241291,
100          0.044999998062849045,
101          0.02499999850988388,
102          0.07000000029802322,
103          0.014999999664723873,
104          0.004999999888241291,
105          0.004999999888241291,
106          0.03999999910593033
107        ],
108        "n": 8
109      }
110    },
111    {
112      "cfg": {
113        "lr": 0.003
114      },
115      "result": {
116        "mean": 0.006874999846331775,
117        "std": 0.006584783448351579,
118        "per_seed": [
119          0.0,
120          0.004999999888241291,
121          0.014999999664723873,
122          0.014999999664723873,
123          0.0,
124          0.004999999888241291,
125          0.0,
126          0.014999999664723873
127        ],
128        "n": 8
129      }
130    },
131    {
132      "cfg": {
133        "lr": 0.01
134      },
135      "result": {
136        "mean": 0.0,
137        "std": 0.0,
138        "per_seed": [
139          0.0,
140          0.0,
141          0.0,
142          0.0,
143          0.0,
144          0.0,
145          0.0,
146          0.0
147        ],
148        "n": 8
149      }
150    }
151  ],
152  "mechanism_signature": {
153    "prediction": "trained fingerprint outputs separate topology pair more than sum baseline",
154    "observed_baseline_output_distance": 0.0020921695977449417,
155    "observed_idea_output_distance": 0.010290324687957764,
156    "predicted_ratio_lower_bound": 1.0,
157    "observed_ratio": 4.918494513565848,
158    "confirmed": true
159  }
160}