Routh-Hurwitz Gain-Capped Optimizer / bench_results.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
  1{
  2  "root_check": [
  3    {
  4      "chi": 0.8,
  5      "max_real_root": 0.0330863273564016
  6    },
  7    {
  8      "chi": 1.0,
  9      "max_real_root": 0.07718936960152892
 10    },
 11    {
 12      "chi": 1.2,
 13      "max_real_root": 0.11566900160083154
 14    }
 15  ],
 16  "bench_report": {
 17    "bench_version": 1,
 18    "track": "dynamics",
 19    "model": "rnn_small",
 20    "metric_direction": "lower is better",
 21    "n_seeds": 8,
 22    "baseline": {
 23      "best_cfg": {
 24        "lr": 0.01
 25      },
 26      "sweep": [
 27        {
 28          "cfg": {
 29            "lr": 0.001
 30          },
 31          "mean": 0.42837614566087723
 32        },
 33        {
 34          "cfg": {
 35            "lr": 0.003
 36          },
 37          "mean": 0.05120387952774763
 38        },
 39        {
 40          "cfg": {
 41            "lr": 0.01
 42          },
 43          "mean": 0.0024460060812998563
 44        }
 45      ],
 46      "full": {
 47        "mean": 0.0025610758020775393,
 48        "std": 0.0009777445230047094,
 49        "per_seed": [
 50          0.004299916326999664,
 51          0.002046577399596572,
 52          0.0017879348015412688,
 53          0.0016495957970619202,
 54          0.0030561184976249933,
 55          0.003266895655542612,
 56          0.0031689393799751997,
 57          0.0012126285582780838
 58        ],
 59        "n": 8
 60      }
 61    },
 62    "idea": {
 63      "per_seed": [
 64        0.004299916326999664,
 65        0.002046577399596572,
 66        0.0017879348015412688,
 67        0.0016495957970619202,
 68        0.0030561184976249933,
 69        0.003266895655542612,
 70        0.0031689393799751997,
 71        0.0012126285582780838
 72      ]
 73    },
 74    "comparison": {
 75      "delta_mean": 0.0,
 76      "idea_wins": 0,
 77      "n_pairs": 8,
 78      "per_seed_diffs": [
 79        0.0,
 80        0.0,
 81        0.0,
 82        0.0,
 83        0.0,
 84        0.0,
 85        0.0,
 86        0.0
 87      ],
 88      "p_value": 1.0,
 89      "mde": 0.0,
 90      "mde_rel_pct": 0.0,
 91      "verdict": "no measurable effect",
 92      "system_worked": false
 93    },
 94    "mechanism_signature": {
 95      "prediction": "online RH cap enforces chi <= rho=0.8",
 96      "predicted_max_chi": 0.8,
 97      "observed_max_chi_idea": 3.605769230769231e-06,
 98      "observed_max_chi_baseline": NaN,
 99      "observed_mean_gain_idea": 1.0000000000000002e-06,
100      "confirmed": true
101    }
102  },
103  "idea_settings": {
104    "0.001": [
105      {
106        "metric": 0.3440668284893036,
107        "history": [
108          0.6993419528007507,
109          0.6813979518413543,
110          0.650027756690979,
111          0.6101660060882569,
112          0.5640302371978759,
113          0.5151371550559998,
114          0.465482257604599,
115          0.4180481779575348
116        ],
117        "max_chi": 3.605769230769231e-07,
118        "mean_effective_lr": 0.001,
119        "mean_observed_gain": 1.0000000000000001e-07,
120        "max_curvature": 0.0001
121      },
122      {
123        "metric": 0.4274647533893585,
124        "history": [
125          0.7602336311340332,
126          0.7395647692680359,
127          0.7040573501586914,
128          0.66268874168396,
129          0.6158934926986694,
130          0.5678855276107788,
131          0.5177900528907776,
132          0.4694719660282135
133        ],
134        "max_chi": 3.605769230769231e-07,
135        "mean_effective_lr": 0.001,
136        "mean_observed_gain": 1.0000000000000001e-07,
137        "max_curvature": 0.0001
138      },
139      {
140        "metric": 0.4056340754032135,
141        "history": [
142          0.6523151779174805,
143          0.6336322736740112,
144          0.5997181987762451,
145          0.5587008571624756,
146          0.5148982381820679,
147          0.47241021394729615,
148          0.4275295877456665,
149          0.38603972673416137
150        ],
151        "max_chi": 3.605769230769231e-07,
152        "mean_effective_lr": 0.001,
153        "mean_observed_gain": 1.0000000000000001e-07,
154        "max_curvature": 0.0001
155      },
156      {
157        "metric": 0.5363389253616333,
158        "history": [
159          0.7145644426345825,
160          0.6977586770057678,
161          0.6691723155975342,
162          0.631492304801941,
163          0.5912015914916993,
164          0.5492246544361115,
165          0.5076988697052002,
166          0.4674421977996826
167        ],
168        "max_chi": 3.605769230769231e-07,
169        "mean_effective_lr": 0.001,
170        "mean_observed_gain": 1.0000000000000001e-07,
171        "max_curvature": 0.0001
172      },
173      {
174        "metric": 0.33050650358200073,
175        "history": [
176          0.6300368976593017,
177          0.6130261039733886,
178          0.5824478936195373,
179          0.5431585824489593,
180          0.49994396686553955,
181          0.4560388541221619,
182          0.4114214777946472,
183          0.3689022183418274
184        ],
185        "max_chi": 3.605769230769231e-07,
186        "mean_effective_lr": 0.001,
187        "mean_observed_gain": 1.0000000000000001e-07,
188        "max_curvature": 0.0001
189      },
190      {
191        "metric": 0.4261091947555542,
192        "history": [
193          0.7953370833396911,
194          0.7722095775604249,
195          0.7340326881408692,
196          0.684270613193512,
197          0.6333953094482422,
198          0.5830701923370362,
199          0.5291142201423645,
200          0.4788055396080017
201        ],
202        "max_chi": 3.605769230769231e-07,
203        "mean_effective_lr": 0.001,
204        "mean_observed_gain": 1.0000000000000001e-07,
205        "max_curvature": 0.0001
206      },
207      {
208        "metric": 0.37957045435905457,
209        "history": [
210          0.7479315209388733,
211          0.7267511749267578,
212          0.6888936424255371,
213          0.6428964591026306,
214          0.5938624811172485,
215          0.544820841550827,
216          0.4944131255149841,
217          0.4452679681777954
218        ],
219        "max_chi": 3.605769230769231e-07,
220        "mean_effective_lr": 0.001,
221        "mean_observed_gain": 1.0000000000000001e-07,
222        "max_curvature": 0.0001
223      },
224      {
225        "metric": 0.4494275450706482,
226        "history": [
227          0.7529299116134643,
228          0.7330988574028016,
229          0.699141960144043,
230          0.6572146892547608,
231          0.6124954748153687,
232          0.5655594277381897,
233          0.5191149485111236,
234          0.4722910523414612
235        ],
236        "max_chi": 3.605769230769231e-07,
237        "mean_effective_lr": 0.001,
238        "mean_observed_gain": 1.0000000000000001e-07,
239        "max_curvature": 0.0001
240      }
241    ],
242    "0.003": [
243      {
244        "metric": 0.018445655703544617,
245        "history": [
246          0.694959774017334,
247          0.6420108795166015,
248          0.5524764275550842,
249          0.44393115520477294,
250          0.3260509085655212,
251          0.21360599756240845,
252          0.11821861624717712,
253          0.052244496941566465
254        ],
255        "max_chi": 1.0817307692307691e-06,
256        "mean_effective_lr": 0.003,
257        "mean_observed_gain": 3.0000000000000004e-07,
258        "max_curvature": 0.0001
259      },
260      {
261        "metric": 0.051231302320957184,
262        "history": [
263          0.7553361463546753,
264          0.6949287056922913,
265          0.5964539051055908,
266          0.48899065852165224,
267          0.375873681306839,
268          0.26908132374286653,
269          0.17072125911712646,
270          0.09357141584157944
271        ],
272        "max_chi": 1.0817307692307691e-06,
273        "mean_effective_lr": 0.003,
274        "mean_observed_gain": 3.0000000000000004e-07,
275        "max_curvature": 0.0001
276      },
277      {
278        "metric": 0.04099540039896965,
279        "history": [
280          0.6478597497940064,
281          0.5935505867004395,
282          0.5012018966674805,
283          0.39846776485443114,
284          0.29653573989868165,
285          0.20853621542453765,
286          0.12716419875621796,
287          0.06803727239370345
288        ],
289        "max_chi": 1.0817307692307691e-06,
290        "mean_effective_lr": 0.003,
291        "mean_observed_gain": 3.0000000000000004e-07,
292        "max_curvature": 0.0001
293      },
294      {
295        "metric": 0.09414315968751907,
296        "history": [
297          0.7108429098129272,
298          0.6615033912658691,
299          0.5811297416687011,
300          0.4809538173675537,
301          0.3801296353340149,
302          0.28171004056930543,
303          0.19256000876426696,
304          0.11847457110881805
305        ],
306        "max_chi": 1.0817307692307691e-06,
307        "mean_effective_lr": 0.003,
308        "mean_observed_gain": 3.0000000000000004e-07,
309        "max_curvature": 0.0001
310      },
311      {
312        "metric": 0.020061630755662918,
313        "history": [
314          0.626112835407257,
315          0.5758749866485595,
316          0.4887616550922394,
317          0.3829389560222626,
318          0.2756062078475952,
319          0.17953147500753402,
320          0.10008062392473221,
321          0.04575570151209831
322        ],
323        "max_chi": 1.0817307692307691e-06,
324        "mean_effective_lr": 0.003,
325        "mean_observed_gain": 3.0000000000000004e-07,
326        "max_curvature": 0.0001
327      },
328      {
329        "metric": 0.03951025754213333,
330        "history": [
331          0.7898440504074097,
332          0.7223627233505249,
333          0.616898854970932,
334          0.4893602728843689,
335          0.3690385365486145,
336          0.25974723160266877,
337          0.1579497426748276,
338          0.08233777940273285
339        ],
340        "max_chi": 1.0817307692307691e-06,
341        "mean_effective_lr": 0.003,
342        "mean_observed_gain": 3.0000000000000004e-07,
343        "max_curvature": 0.0001
344      },
345      {
346        "metric": 0.031925592571496964,
347        "history": [
348          0.7429254531860352,
349          0.6810082459449768,
350          0.5761089265346527,
351          0.4574007022380829,
352          0.3400264596939087,
353          0.2332361841201782,
354          0.13982866019010542,
355          0.06969675615429878
356        ],
357        "max_chi": 1.0817307692307691e-06,
358        "mean_effective_lr": 0.003,
359        "mean_observed_gain": 3.0000000000000004e-07,
360        "max_curvature": 0.0001
361      },
362      {
363        "metric": 0.05982330068945885,
364        "history": [
365          0.7481838035583496,
366          0.6902918517589569,
367          0.5962544202804565,
368          0.4881104326248169,
369          0.3803297543525696,
370          0.27569568634033204,
371          0.18162715554237366,
372          0.10307565599679946
373        ],
374        "max_chi": 1.0817307692307691e-06,
375        "mean_effective_lr": 0.003,
376        "mean_observed_gain": 3.0000000000000004e-07,
377        "max_curvature": 0.0001
378      }
379    ],
380    "0.01": [
381      {
382        "metric": 0.004299916326999664,
383        "history": [
384          0.6798131060600281,
385          0.512830947637558,
386          0.26385428369045255,
387          0.055778834596276285,
388          0.017274612039327623,
389          0.07015145182609558,
390          0.0401259084790945,
391          0.0036679642274975778
392        ],
393        "max_chi": 3.605769230769231e-06,
394        "mean_effective_lr": 0.01,
395        "mean_observed_gain": 1.0000000000000002e-06,
396        "max_curvature": 0.0001
397      },
398      {
399        "metric": 0.002046577399596572,
400        "history": [
401          0.7385136270523072,
402          0.5534801185131073,
403          0.2967984753847122,
404          0.08821958027780057,
405          0.009234225079417228,
406          0.047636431157588956,
407          0.04728898391127587,
408          0.010239683538675309
409        ],
410        "max_chi": 3.605769230769231e-06,
411        "mean_effective_lr": 0.01,
412        "mean_observed_gain": 1.0000000000000002e-06,
413        "max_curvature": 0.0001
414      },
415      {
416        "metric": 0.0017879348015412688,
417        "history": [
418          0.6325315642356872,
419          0.469499454498291,
420          0.2445989990234375,
421          0.06959607563912869,
422          0.013648657500743866,
423          0.05180964022874832,
424          0.043835976719856264,
425          0.008696820829063654
426        ],
427        "max_chi": 3.605769230769231e-06,
428        "mean_effective_lr": 0.01,
429        "mean_observed_gain": 1.0000000000000002e-06,
430        "max_curvature": 0.0001
431      },
432      {
433        "metric": 0.0016495957970619202,
434        "history": [
435          0.6980253458023071,
436          0.544585394859314,
437          0.32391603231430055,
438          0.10666814312338829,
439          0.00840164065361023,
440          0.041370466351509094,
441          0.06321462988853455,
442          0.022033566143363714
443        ],
444        "max_chi": 3.605769230769231e-06,
445        "mean_effective_lr": 0.01,
446        "mean_observed_gain": 1.0000000000000002e-06,
447        "max_curvature": 0.0001
448      },
449      {
450        "metric": 0.0030561184976249933,
451        "history": [
452          0.6125363111495972,
453          0.4540520048141479,
454          0.21791321754455567,
455          0.036135574914515016,
456          0.023327344805002214,
457          0.06946323305368424,
458          0.03671337585896253,
459          0.0036678536795079707
460        ],
461        "max_chi": 3.605769230769231e-06,
462        "mean_effective_lr": 0.01,
463        "mean_observed_gain": 1.0000000000000002e-06,
464        "max_curvature": 0.0001
465      },
466      {
467        "metric": 0.003266895655542612,
468        "history": [
469          0.7710019874572754,
470          0.5653183555603027,
471          0.29373787581920624,
472          0.06810168888419867,
473          0.017942358553409577,
474          0.058926407247781754,
475          0.041252076476812366,
476          0.006175353517755866
477        ],
478        "max_chi": 3.605769230769231e-06,
479        "mean_effective_lr": 0.01,
480        "mean_observed_gain": 1.0000000000000002e-06,
481        "max_curvature": 0.0001
482      },
483      {
484        "metric": 0.0031689393799751997,
485        "history": [
486          0.7256936359405518,
487          0.5360211396217346,
488          0.2658588397502899,
489          0.05989395052194595,
490          0.02104116842150688,
491          0.06888968884944915,
492          0.047494716197252276,
493          0.007844956731423736
494        ],
495        "max_chi": 3.605769230769231e-06,
496        "mean_effective_lr": 0.01,
497        "mean_observed_gain": 1.0000000000000002e-06,
498        "max_curvature": 0.0001
499      },
500      {
501        "metric": 0.0012126285582780838,
502        "history": [
503          0.7319039154052734,
504          0.554954423904419,
505          0.3081369322538376,
506          0.09451350867748261,
507          0.007374951243400573,
508          0.04413643956184387,
509          0.05444020733237267,
510          0.015588753465563059
511        ],
512        "max_chi": 3.605769230769231e-06,
513        "mean_effective_lr": 0.01,
514        "mean_observed_gain": 1.0000000000000002e-06,
515        "max_curvature": 0.0001
516      }
517    ]
518  },
519  "custom_track": null
520}