Dual-Ensemble Latent Transition Model / 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": 0.0
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 0.004298779065720737
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.003,
 23          "weight_decay": 0.0
 24        },
 25        "mean": 0.0034615940821822733
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.01,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 0.0014865802077110857
 33      }
 34    ],
 35    "full": {
 36      "mean": 0.0014198302378645167,
 37      "std": 0.00039839114901047844,
 38      "per_seed": [
 39        0.0009327387670055032,
 40        0.0018641706556081772,
 41        0.0010785290505737066,
 42        0.0020708823576569557,
 43        0.0009260809747502208,
 44        0.0014055331703275442,
 45        0.0016443432541564107,
 46        0.001436363672837615
 47      ],
 48      "n": 8
 49    }
 50  },
 51  "idea": {
 52    "mean": 0.001362547576718498,
 53    "std": 0.00047644182050365723,
 54    "per_seed": [
 55      0.0014115098165348172,
 56      0.0007051810971461236,
 57      0.0013796384446322918,
 58      0.001993401674553752,
 59      0.001138740568421781,
 60      0.0022058519534766674,
 61      0.0010818139417096972,
 62      0.0009842431172728539
 63    ],
 64    "n": 8
 65  },
 66  "comparison": {
 67    "delta_mean": -5.7282661146018654e-05,
 68    "idea_wins": 4,
 69    "n_pairs": 8,
 70    "per_seed_diffs": [
 71      0.00047877104952931404,
 72      -0.0011589895584620535,
 73      0.00030110939405858517,
 74      -7.748068310320377e-05,
 75      0.0002126595936715603,
 76      0.0008003187831491232,
 77      -0.0005625293124467134,
 78      -0.00045212055556476116
 79    ],
 80    "p_value": 0.7975,
 81    "mde": 0.0005334826450520739,
 82    "mde_rel_pct": 37.573692320742076,
 83    "verdict": "no significant win",
 84    "system_worked": false
 85  },
 86  "mechanism_signature": {
 87    "quantity": "source-to-sink subset test MSE (trained models)",
 88    "baseline_predicted": 0.0010878387474804185,
 89    "dual_predicted": 0.0006060761043045204,
 90    "observed_reduction": 0.0004817626431758981,
 91    "prediction": "Q should reduce directed source-sink prediction error",
 92    "confirmed": true
 93  },
 94  "idea_sweep": [
 95    {
 96      "cfg": {
 97        "lr": 0.001,
 98        "weight_decay": 0.0
 99      },
100      "full": {
101        "mean": 0.03154140734113753,
102        "std": 0.0061376140406840096,
103        "per_seed": [
104          0.03103465959429741,
105          0.03406568616628647,
106          0.028227416798472404,
107          0.033552005887031555,
108          0.03155306726694107,
109          0.04121524840593338,
110          0.01821603998541832,
111          0.03446713462471962
112        ],
113        "n": 8
114      }
115    },
116    {
117      "cfg": {
118        "lr": 0.003,
119        "weight_decay": 0.0
120      },
121      "full": {
122        "mean": 0.005806407978525385,
123        "std": 0.0014531884675848988,
124        "per_seed": [
125          0.005496995989233255,
126          0.0035544722341001034,
127          0.006204786244779825,
128          0.008257976733148098,
129          0.0056442818604409695,
130          0.003874243935570121,
131          0.006419075652956963,
132          0.006999431177973747
133        ],
134        "n": 8
135      }
136    },
137    {
138      "cfg": {
139        "lr": 0.01,
140        "weight_decay": 0.0
141      },
142      "full": {
143        "mean": 0.001362547576718498,
144        "std": 0.00047644182050365723,
145        "per_seed": [
146          0.0014115098165348172,
147          0.0007051810971461236,
148          0.0013796384446322918,
149          0.001993401674553752,
150          0.001138740568421781,
151          0.0022058519534766674,
152          0.0010818139417096972,
153          0.0009842431172728539
154        ],
155        "n": 8
156      }
157    }
158  ],
159  "custom_track": null
160}