Wasserstein Speed-Limit Controller / bench_report.json

Failed on benchmark

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "dynamics",
  4  "model": "rnn_small",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.01,
 10      "weight_decay": 0.0001
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 0.0035823711077682674
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.001,
 23          "weight_decay": 0.0001
 24        },
 25        "mean": 0.0035786936641670763
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.003,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 0.0033184013445861638
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "weight_decay": 0.0001
 38        },
 39        "mean": 0.0032864113454706967
 40      },
 41      {
 42        "cfg": {
 43          "lr": 0.01,
 44          "weight_decay": 0.0
 45        },
 46        "mean": 0.0015948790241964161
 47      },
 48      {
 49        "cfg": {
 50          "lr": 0.01,
 51          "weight_decay": 0.0001
 52        },
 53        "mean": 0.0015534490812569857
 54      }
 55    ],
 56    "full": {
 57      "mean": 0.0013736290202359669,
 58      "std": 0.0004336284119124816,
 59      "per_seed": [
 60        0.0022530313581228256,
 61        0.0016054200241342187,
 62        0.0008981260471045971,
 63        0.0014572188956663013,
 64        0.0016656755469739437,
 65        0.0009697971981950104,
 66        0.000995026552118361,
 67        0.0011447365395724773
 68      ],
 69      "n": 8
 70    }
 71  },
 72  "idea": {
 73    "mean": 0.4975068047642708,
 74    "std": 0.06142691120766916,
 75    "per_seed": [
 76      0.39669302105903625,
 77      0.551139771938324,
 78      0.4963032305240631,
 79      0.5875324010848999,
 80      0.44322699308395386,
 81      0.4858019948005676,
 82      0.45647817850112915,
 83      0.5628788471221924
 84    ],
 85    "n": 8
 86  },
 87  "comparison": {
 88    "delta_mean": 0.4961331757440348,
 89    "idea_wins": 0,
 90    "n_pairs": 8,
 91    "per_seed_diffs": [
 92      0.39443998970091343,
 93      0.5495343519141898,
 94      0.4954051044769585,
 95      0.5860751821892336,
 96      0.4415613175369799,
 97      0.4848321976023726,
 98      0.4554831519490108,
 99      0.5617341105826199
100    ],
101    "p_value": 0.0081,
102    "mde": 0.055052955593549176,
103    "mde_rel_pct": 4007.847445163323,
104    "verdict": "idea worse (significant)",
105    "system_worked": false
106  },
107  "mechanism_signature": {
108    "signature": {
109      "quantity": "W2^2/(D*dt*Sigma)",
110      "predicted": "<= 1 in stable intervals",
111      "observed_mean_ratio": 12911.468078824939,
112      "observed_max_ratio_mean": 18444.135724120242,
113      "observed_final_eta_mean": 0.0025,
114      "confirmed": false
115    },
116    "idea_grid": [
117      0.001,
118      0.003,
119      0.01
120    ],
121    "baseline_grid": [
122      {
123        "lr": 0.001,
124        "weight_decay": 0.0
125      },
126      {
127        "lr": 0.001,
128        "weight_decay": 0.0001
129      },
130      {
131        "lr": 0.003,
132        "weight_decay": 0.0
133      },
134      {
135        "lr": 0.003,
136        "weight_decay": 0.0001
137      },
138      {
139        "lr": 0.01,
140        "weight_decay": 0.0
141      },
142      {
143        "lr": 0.01,
144        "weight_decay": 0.0001
145      }
146    ],
147    "protocol": "8 paired seeds; 4-seed baseline/idea selection"
148  },
149  "idea_configs_considered": [
150    {
151      "lr": 0.001
152    },
153    {
154      "lr": 0.003
155    },
156    {
157      "lr": 0.01
158    }
159  ]
160}