Lyapunov-Calibrated Multiplicative Noise / 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": 7,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006,
 10      "knob": 0.01
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "knob": 0.01
 17        },
 18        "mean": 0.01457586046308279
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.001,
 23          "knob": 0.03
 24        },
 25        "mean": 0.01469542644917965
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.001,
 30          "knob": 0.06
 31        },
 32        "mean": 0.0148483463563025
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "knob": 0.01
 38        },
 39        "mean": 0.0047323108301498
 40      },
 41      {
 42        "cfg": {
 43          "lr": 0.003,
 44          "knob": 0.03
 45        },
 46        "mean": 0.00487289740704
 47      },
 48      {
 49        "cfg": {
 50          "lr": 0.003,
 51          "knob": 0.06
 52        },
 53        "mean": 0.00508440809790045
 54      },
 55      {
 56        "cfg": {
 57          "lr": 0.006,
 58          "knob": 0.01
 59        },
 60        "mean": 0.0026383594959042966
 61      },
 62      {
 63        "cfg": {
 64          "lr": 0.006,
 65          "knob": 0.03
 66        },
 67        "mean": 0.002871918957680464
 68      },
 69      {
 70        "cfg": {
 71          "lr": 0.006,
 72          "knob": 0.06
 73        },
 74        "mean": 0.0032478051725775003
 75      }
 76    ],
 77    "full": {
 78      "mean": 0.002745348698226735,
 79      "std": 0.0005053976729980165,
 80      "per_seed": [
 81        0.0029019152279943228,
 82        0.002326412359252572,
 83        0.00219854642637074,
 84        0.0031265639699995518,
 85        0.0031316187232732773,
 86        0.0024797292426228523,
 87        0.0021466391626745462,
 88        0.0036513644736260176
 89      ],
 90      "n": 8
 91    }
 92  },
 93  "idea": {
 94    "mean": 0.0027424886024424006,
 95    "std": 0.0004920238808410106,
 96    "per_seed": [
 97      0.002906925044953823,
 98      0.002241562819108367,
 99      0.0021613340359181166,
100      0.003099968424066901,
101      0.0030245676171034575,
102      0.002236209576949477,
103      0.003526852698996663
104    ],
105    "n": 7
106  },
107  "comparison": {
108    "delta_mean": 0.00012657072927270616,
109    "idea_wins": 5,
110    "n_pairs": 7,
111    "per_seed_diffs": [
112      5.009816959500313e-06,
113      -8.48495401442051e-05,
114      -3.721239045262337e-05,
115      -2.6595545932650566e-05,
116      -0.00010705110616981983,
117      -0.00024351966567337513,
118      0.0013802135363221169
119    ],
120    "p_value": 0.98495,
121    "mde": 0.0004994211111627109,
122    "mde_rel_pct": 18.191536524496687,
123    "verdict": "no significant win",
124    "system_worked": false
125  },
126  "mechanism_signature": {
127    "idea_sweep": [
128      {
129        "cfg": {
130          "lr": 0.001,
131          "knob": 0.35
132        },
133        "mean": 0.014552626991644502
134      },
135      {
136        "cfg": {
137          "lr": 0.001,
138          "knob": 0.7
139        },
140        "mean": 0.014597104163840413
141      },
142      {
143        "cfg": {
144          "lr": 0.001,
145          "knob": 1.4
146        },
147        "mean": 0.014682337641716003
148      },
149      {
150        "cfg": {
151          "lr": 0.003,
152          "knob": 0.35
153        },
154        "mean": 0.004710524983238429
155      },
156      {
157        "cfg": {
158          "lr": 0.003,
159          "knob": 0.7
160        },
161        "mean": 0.004761852149385959
162      },
163      {
164        "cfg": {
165          "lr": 0.003,
166          "knob": 1.4
167        },
168        "mean": 0.004872480174526572
169      },
170      {
171        "cfg": {
172          "lr": 0.006,
173          "knob": 0.35
174        },
175        "mean": 0.002602447581011802
176      },
177      {
178        "cfg": {
179          "lr": 0.006,
180          "knob": 0.7
181        },
182        "mean": 0.002704756858292967
183      },
184      {
185        "cfg": {
186          "lr": 0.006,
187          "knob": 1.4
188        },
189        "mean": 0.0029094331548549235
190      }
191    ],
192    "mechanism_signature": {
193      "trained_model": "ResidualRNN dynamics forecaster",
194      "n_observations": 56,
195      "predicted_conditional_gate_variance": "q*u*(1-u), q=2*kappa*dt*[r-r_target]+",
196      "observed_variance_fit_slope": 1.0019951817912731,
197      "expected_slope": 1.0,
198      "relative_error": 0.001995181791273115,
199      "confirmed": true
200    }
201  },
202  "custom_track": null
203}