Entropy-calibrated hyperbolic curvature / bench_report.json

Mechanism confirmed, baseline not beaten

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_config": {
  9      "lr": 0.006,
 10      "epochs": 12
 11    },
 12    "sweep": {
 13      "0.001": 0.004298779065720737,
 14      "0.003": 0.0034615940821822733,
 15      "0.006": 0.0018282315868418664
 16    },
 17    "full": {
 18      "per_seed": [
 19        0.002004395006224513,
 20        0.001437882543541491,
 21        0.0015748648438602686,
 22        0.002295783953741193,
 23        0.0024020641576498747,
 24        0.0024898042902350426,
 25        0.0020851334556937218,
 26        0.0025566271506249905
 27      ]
 28    },
 29    "all_configs": {
 30      "0.001": [
 31        0.007754037156701088,
 32        0.003745198715478182,
 33        0.0025475025177001953,
 34        0.003148377873003483,
 35        0.004954223055392504,
 36        0.004425588063895702,
 37        0.0044244141317903996,
 38        0.009162315167486668
 39      ],
 40      "0.003": [
 41        0.002470282604917884,
 42        0.002755755092948675,
 43        0.0011578478151932359,
 44        0.007462490815669298,
 45        0.0025412666145712137,
 46        0.0030948929488658905,
 47        0.0021244161762297153,
 48        0.0019843443296849728
 49      ],
 50      "0.006": [
 51        0.002004395006224513,
 52        0.001437882543541491,
 53        0.0015748648438602686,
 54        0.002295783953741193,
 55        0.0024020641576498747,
 56        0.0024898042902350426,
 57        0.0020851334556937218,
 58        0.0025566271506249905
 59      ]
 60    }
 61  },
 62  "idea": {
 63    "best_config": {
 64      "lr": 0.006,
 65      "epochs": 12
 66    },
 67    "sweep_means": {
 68      "0.001": 0.0045381716336123645,
 69      "0.003": 0.0035767197841778398,
 70      "0.006": 0.0014118490216787905
 71    },
 72    "per_seed": [
 73      0.0018144856439903378,
 74      0.0014102724380791187,
 75      0.0007331487722694874,
 76      0.0016894892323762178,
 77      0.0015310372691601515,
 78      0.0012987027876079082,
 79      0.002770845778286457,
 80      0.0015978770097717643
 81    ],
 82    "signature_samples": [
 83      {
 84        "entropy": 0.7434267299673585,
 85        "curvature": 0.056044551169079165,
 86        "radius_mean": 1.352136492729187,
 87        "radius_tau": 0.3528751730918884
 88      },
 89      {
 90        "entropy": 0.8414160177487802,
 91        "curvature": 0.056044551169079165,
 92        "radius_mean": 1.300600290298462,
 93        "radius_tau": 0.3686612844467163
 94      },
 95      {
 96        "entropy": 0.8015473816323408,
 97        "curvature": 0.056044551169079165,
 98        "radius_mean": 1.31516432762146,
 99        "radius_tau": 0.41591450572013855
100      },
101      {
102        "entropy": 0.7520167144600518,
103        "curvature": 0.056044551169079165,
104        "radius_mean": 1.4210695028305054,
105        "radius_tau": 0.40408262610435486
106      },
107      {
108        "entropy": 0.8041456100108358,
109        "curvature": 0.056044551169079165,
110        "radius_mean": 1.2326090335845947,
111        "radius_tau": 0.40702223777770996
112      },
113      {
114        "entropy": 0.7911075937989723,
115        "curvature": 0.056044551169079165,
116        "radius_mean": 1.3452383279800415,
117        "radius_tau": 0.3651469051837921
118      },
119      {
120        "entropy": 0.8041456100108358,
121        "curvature": 0.056044551169079165,
122        "radius_mean": 1.3425686359405518,
123        "radius_tau": 0.41605350375175476
124      },
125      {
126        "entropy": 0.7550113436280443,
127        "curvature": 0.056044551169079165,
128        "radius_mean": 1.4108587503433228,
129        "radius_tau": 0.4241693317890167
130      }
131    ]
132  },
133  "comparison": {
134    "delta_mean": -0.0005000870587537065,
135    "idea_wins": 7,
136    "n_pairs": 8,
137    "per_seed_diffs": [
138      -0.0001899093622341752,
139      -2.7610105462372303e-05,
140      -0.0008417160715907812,
141      -0.000606294721364975,
142      -0.0008710268884897232,
143      -0.0011911015026271343,
144      0.0006857123225927353,
145      -0.0009587501408532262
146    ],
147    "p_value": 0.05505,
148    "mde": 0.0005172797403971514,
149    "mde_rel_pct": 24.56429711910889,
150    "verdict": "no significant win",
151    "system_worked": false
152  },
153  "math_sanity": {
154    "entropy_decreasing_fraction": 1.0,
155    "lambda_true": 1.1249358995152312,
156    "lambda_recovered": 1.1249358995152312,
157    "amplification": 1.657187089473768
158  },
159  "mechanism_signature": {
160    "predicted_lambda": 1.1249358995152312,
161    "observed_entropy_curve_recovered_lambda": 1.1249358995152312,
162    "trained_model_entropy_mean": 0.7866021251571524,
163    "trained_model_curvature_mean": 0.056044551169079165,
164    "confirmed": true
165  }
166}