Distinct-kink complexity regularizer and merger / results/bench_report.json

Failed on benchmark

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "tabular",
  4  "model": "mlp_tiny",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.01,
 10      "weight_decay": 0.0
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 24.645838737487793
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.003,
 23          "weight_decay": 0.0
 24        },
 25        "mean": 13.62548279762268
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.01,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 8.338012218475342
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "weight_decay": 0.0001
 38        },
 39        "mean": 13.600164413452148
 40      }
 41    ],
 42    "full": {
 43      "mean": 8.617289543151855,
 44      "std": 0.739608681399212,
 45      "per_seed": [
 46        8.399882316589355,
 47        7.785224437713623,
 48        7.834766864776611,
 49        9.332175254821777,
 50        7.886314392089844,
 51        8.663776397705078,
 52        9.938685417175293,
 53        9.097491264343262
 54      ],
 55      "n": 8
 56    }
 57  },
 58  "idea": {
 59    "mean": 8.601594865322113,
 60    "std": 0.7732046694513341,
 61    "per_seed": [
 62      8.518109321594238,
 63      7.764858722686768,
 64      7.801291465759277,
 65      9.256509780883789,
 66      7.880615234375,
 67      8.263826370239258,
 68      9.904521942138672,
 69      9.423026084899902
 70    ],
 71    "n": 8
 72  },
 73  "comparison": {
 74    "delta_mean": -0.01569467782974243,
 75    "idea_wins": 6,
 76    "n_pairs": 8,
 77    "per_seed_diffs": [
 78      0.11822700500488281,
 79      -0.02036571502685547,
 80      -0.033475399017333984,
 81      -0.07566547393798828,
 82      -0.00569915771484375,
 83      -0.3999500274658203,
 84      -0.034163475036621094,
 85      0.3255348205566406
 86    ],
 87    "p_value": 0.75635,
 88    "mde": 0.1691312343107815,
 89    "mde_rel_pct": 1.9626964309814772,
 90    "verdict": "no measurable effect",
 91    "system_worked": false
 92  },
 93  "idea_sweep": [
 94    {
 95      "cfg": {
 96        "lr": 0.001,
 97        "lambda_k": 0.0001
 98      },
 99      "mean": 23.893144130706787
100    },
101    {
102      "cfg": {
103        "lr": 0.003,
104        "lambda_k": 0.0001
105      },
106      "mean": 13.534956455230713
107    },
108    {
109      "cfg": {
110        "lr": 0.01,
111        "lambda_k": 0.0001
112      },
113      "mean": 8.335192322731018
114    }
115  ],
116  "matched_structure": "MLP/tabular optimizer-regularizer track",
117  "mechanism_signature": {
118    "predicted_duplicate_fraction_mean": 0.0,
119    "observed_duplicate_fraction_mean": 0.0,
120    "probe_prediction_rms_mean": 15.10029935836792,
121    "confirmed": false
122  }
123}