Differentiable Maximal-Attractor Trap / bench_report.json
Mechanism confirmed, baseline not beaten
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 "epochs": 12,
11 "trap_lambda": 0.0
12 },
13 "sweep": [
14 {
15 "cfg": {
16 "lr": 0.001,
17 "epochs": 12,
18 "trap_lambda": 0.0
19 },
20 "per_seed": [
21 0.02792397141456604,
22 0.041552960872650146,
23 0.03038809262216091,
24 0.03133251145482063,
25 0.05597708746790886,
26 0.022259902209043503,
27 0.07735025882720947,
28 0.01893964223563671
29 ],
30 "mean": 0.038215553387999535,
31 "std": 0.019635395061371623
32 },
33 {
34 "cfg": {
35 "lr": 0.003,
36 "epochs": 12,
37 "trap_lambda": 0.0
38 },
39 "per_seed": [
40 0.009584075771272182,
41 0.013722526840865612,
42 0.008015970699489117,
43 0.011999157257378101,
44 0.01405302993953228,
45 0.00778926583006978,
46 0.013168145902454853,
47 0.007831467315554619
48 ],
49 "mean": 0.010770454944577068,
50 "std": 0.0027585833372694973
51 },
52 {
53 "cfg": {
54 "lr": 0.01,
55 "epochs": 12,
56 "trap_lambda": 0.0
57 },
58 "per_seed": [
59 0.0024895062670111656,
60 0.0038835888262838125,
61 0.003609295468777418,
62 0.003999150358140469,
63 0.0032447325065732002,
64 0.0038587753660976887,
65 0.008323567919433117,
66 0.0034713787026703358
67 ],
68 "mean": 0.004109999426873401,
69 "std": 0.0017693440291274862
70 },
71 {
72 "cfg": {
73 "lr": 0.001,
74 "epochs": 12,
75 "trap_lambda": 0.3
76 },
77 "per_seed": [
78 0.02792397141456604,
79 0.041552960872650146,
80 0.03038809262216091,
81 0.03133251145482063,
82 0.05597708746790886,
83 0.022259902209043503,
84 0.07735025882720947,
85 0.01893964223563671
86 ],
87 "mean": 0.038215553387999535,
88 "std": 0.019635395061371623
89 },
90 {
91 "cfg": {
92 "lr": 0.003,
93 "epochs": 12,
94 "trap_lambda": 0.3
95 },
96 "per_seed": [
97 0.009584075771272182,
98 0.013722526840865612,
99 0.008015970699489117,
100 0.011999157257378101,
101 0.01405302993953228,
102 0.00778926583006978,
103 0.013168145902454853,
104 0.007831467315554619
105 ],
106 "mean": 0.010770454944577068,
107 "std": 0.0027585833372694973
108 },
109 {
110 "cfg": {
111 "lr": 0.01,
112 "epochs": 12,
113 "trap_lambda": 0.3
114 },
115 "per_seed": [
116 0.0024895062670111656,
117 0.0038835888262838125,
118 0.003609295468777418,
119 0.003999150358140469,
120 0.0032447325065732002,
121 0.0038587753660976887,
122 0.008323567919433117,
123 0.0034713787026703358
124 ],
125 "mean": 0.004109999426873401,
126 "std": 0.0017693440291274862
127 },
128 {
129 "cfg": {
130 "lr": 0.001,
131 "epochs": 12,
132 "trap_lambda": 1.0
133 },
134 "per_seed": [
135 0.02792397141456604,
136 0.041552960872650146,
137 0.03038809262216091,
138 0.03133251145482063,
139 0.05597708746790886,
140 0.022259902209043503,
141 0.07735025882720947,
142 0.01893964223563671
143 ],
144 "mean": 0.038215553387999535,
145 "std": 0.019635395061371623
146 },
147 {
148 "cfg": {
149 "lr": 0.003,
150 "epochs": 12,
151 "trap_lambda": 1.0
152 },
153 "per_seed": [
154 0.009584075771272182,
155 0.013722526840865612,
156 0.008015970699489117,
157 0.011999157257378101,
158 0.01405302993953228,
159 0.00778926583006978,
160 0.013168145902454853,
161 0.007831467315554619
162 ],
163 "mean": 0.010770454944577068,
164 "std": 0.0027585833372694973
165 },
166 {
167 "cfg": {
168 "lr": 0.01,
169 "epochs": 12,
170 "trap_lambda": 1.0
171 },
172 "per_seed": [
173 0.0024895062670111656,
174 0.0038835888262838125,
175 0.003609295468777418,
176 0.003999150358140469,
177 0.0032447325065732002,
178 0.0038587753660976887,
179 0.008323567919433117,
180 0.0034713787026703358
181 ],
182 "mean": 0.004109999426873401,
183 "std": 0.0017693440291274862
184 }
185 ],
186 "harness_tuning": {
187 "best_cfg": {
188 "lr": 0.01,
189 "epochs": 12,
190 "trap_lambda": 0.0
191 },
192 "sweep": [
193 {
194 "cfg": {
195 "lr": 0.001,
196 "epochs": 12,
197 "trap_lambda": 0.0
198 },
199 "mean": 0.03279938409104943
200 },
201 {
202 "cfg": {
203 "lr": 0.003,
204 "epochs": 12,
205 "trap_lambda": 0.0
206 },
207 "mean": 0.010830432642251253
208 },
209 {
210 "cfg": {
211 "lr": 0.01,
212 "epochs": 12,
213 "trap_lambda": 0.0
214 },
215 "mean": 0.0034953852300532162
216 },
217 {
218 "cfg": {
219 "lr": 0.001,
220 "epochs": 12,
221 "trap_lambda": 0.3
222 },
223 "mean": 0.03279938409104943
224 },
225 {
226 "cfg": {
227 "lr": 0.003,
228 "epochs": 12,
229 "trap_lambda": 0.3
230 },
231 "mean": 0.010830432642251253
232 },
233 {
234 "cfg": {
235 "lr": 0.01,
236 "epochs": 12,
237 "trap_lambda": 0.3
238 },
239 "mean": 0.0034953852300532162
240 },
241 {
242 "cfg": {
243 "lr": 0.001,
244 "epochs": 12,
245 "trap_lambda": 1.0
246 },
247 "mean": 0.03279938409104943
248 },
249 {
250 "cfg": {
251 "lr": 0.003,
252 "epochs": 12,
253 "trap_lambda": 1.0
254 },
255 "mean": 0.010830432642251253
256 },
257 {
258 "cfg": {
259 "lr": 0.01,
260 "epochs": 12,
261 "trap_lambda": 1.0
262 },
263 "mean": 0.0034953852300532162
264 }
265 ],
266 "full": {
267 "mean": 0.004109999426873401,
268 "std": 0.0016550697890823066,
269 "per_seed": [
270 0.0024895062670111656,
271 0.0038835888262838125,
272 0.003609295468777418,
273 0.003999150358140469,
274 0.0032447325065732002,
275 0.0038587753660976887,
276 0.008323567919433117,
277 0.0034713787026703358
278 ],
279 "n": 8
280 }
281 },
282 "full": {
283 "cfg": {
284 "lr": 0.01,
285 "epochs": 12,
286 "trap_lambda": 0.0
287 },
288 "per_seed": [
289 0.0024895062670111656,
290 0.0038835888262838125,
291 0.003609295468777418,
292 0.003999150358140469,
293 0.0032447325065732002,
294 0.0038587753660976887,
295 0.008323567919433117,
296 0.0034713787026703358
297 ],
298 "mean": 0.004109999426873401,
299 "std": 0.0017693440291274862
300 }
301 },
302 "idea": {
303 "best_cfg": {
304 "lr": 0.01,
305 "epochs": 12,
306 "trap_lambda": 0.3
307 },
308 "per_seed": [
309 0.0020382918883115053,
310 0.002796109300106764,
311 0.00355388387106359,
312 0.0032515127677470446,
313 0.003949518781155348,
314 0.003303474746644497,
315 0.0037955872248858213,
316 0.004127324093133211
317 ],
318 "mean": 0.0033519628341309726,
319 "std": 0.0006806487150040169
320 },
321 "comparison": {
322 "delta_mean": -0.0007580365927424282,
323 "idea_wins": 6,
324 "n_pairs": 8,
325 "per_seed_diffs": [
326 -0.0004512143786996603,
327 -0.0010874795261770487,
328 -5.541159771382809e-05,
329 -0.000747637590393424,
330 0.0007047862745821476,
331 -0.0005553006194531918,
332 -0.004527980694547296,
333 0.0006559453904628754
334 ],
335 "p_value": 0.22655,
336 "mde": 0.0013808020917505462,
337 "mde_rel_pct": 33.59616263501437,
338 "verdict": "no significant win",
339 "system_worked": false
340 },
341 "mechanism_signature": {
342 "track_choice": "dynamics is structurally matched to recurrent stability",
343 "idea_sweep": [
344 {
345 "cfg": {
346 "lr": 0.001,
347 "epochs": 12,
348 "trap_lambda": 0.0
349 },
350 "per_seed": [
351 0.02792397141456604,
352 0.041552960872650146,
353 0.03038809262216091,
354 0.03133251145482063,
355 0.05597708746790886,
356 0.022259902209043503,
357 0.07735025882720947,
358 0.01893964223563671
359 ],
360 "mean": 0.038215553387999535,
361 "std": 0.019635395061371623
362 },
363 {
364 "cfg": {
365 "lr": 0.003,
366 "epochs": 12,
367 "trap_lambda": 0.0
368 },
369 "per_seed": [
370 0.009584075771272182,
371 0.013722526840865612,
372 0.008015970699489117,
373 0.011999157257378101,
374 0.01405302993953228,
375 0.00778926583006978,
376 0.013168145902454853,
377 0.007831467315554619
378 ],
379 "mean": 0.010770454944577068,
380 "std": 0.0027585833372694973
381 },
382 {
383 "cfg": {
384 "lr": 0.01,
385 "epochs": 12,
386 "trap_lambda": 0.0
387 },
388 "per_seed": [
389 0.0024895062670111656,
390 0.0038835888262838125,
391 0.003609295468777418,
392 0.003999150358140469,
393 0.0032447325065732002,
394 0.0038587753660976887,
395 0.008323567919433117,
396 0.0034713787026703358
397 ],
398 "mean": 0.004109999426873401,
399 "std": 0.0017693440291274862
400 },
401 {
402 "cfg": {
403 "lr": 0.001,
404 "epochs": 12,
405 "trap_lambda": 0.3
406 },
407 "per_seed": [
408 0.029911765828728676,
409 0.043103478848934174,
410 0.029594196006655693,
411 0.03148665651679039,
412 0.056626711040735245,
413 0.023263107985258102,
414 0.07945620268583298,
415 0.019382713362574577
416 ],
417 "mean": 0.03910310403443873,
418 "std": 0.020086107344229626
419 },
420 {
421 "cfg": {
422 "lr": 0.003,
423 "epochs": 12,
424 "trap_lambda": 0.3
425 },
426 "per_seed": [
427 0.00927602220326662,
428 0.01410755980759859,
429 0.008517258800566196,
430 0.011783787980675697,
431 0.014786372892558575,
432 0.0075119370594620705,
433 0.012531742453575134,
434 0.008021011017262936
435 ],
436 "mean": 0.010816961526870728,
437 "std": 0.002850141431686086
438 },
439 {
440 "cfg": {
441 "lr": 0.01,
442 "epochs": 12,
443 "trap_lambda": 0.3
444 },
445 "per_seed": [
446 0.0020382918883115053,
447 0.002796109300106764,
448 0.00355388387106359,
449 0.0032515127677470446,
450 0.003949518781155348,
451 0.003303474746644497,
452 0.0037955872248858213,
453 0.004127324093133211
454 ],
455 "mean": 0.0033519628341309726,
456 "std": 0.0006806487150040169
457 },
458 {
459 "cfg": {
460 "lr": 0.001,
461 "epochs": 12,
462 "trap_lambda": 1.0
463 },
464 "per_seed": [
465 0.02979419380426407,
466 0.04312347620725632,
467 0.02988792397081852,
468 0.0319940410554409,
469 0.05702750012278557,
470 0.02334035187959671,
471 0.07955670356750488,
472 0.019355962052941322
473 ],
474 "mean": 0.039260019082576036,
475 "std": 0.020122028461056908
476 },
477 {
478 "cfg": {
479 "lr": 0.003,
480 "epochs": 12,
481 "trap_lambda": 1.0
482 },
483 "per_seed": [
484 0.009333031252026558,
485 0.014507703483104706,
486 0.00847854744642973,
487 0.012289805337786674,
488 0.015088483691215515,
489 0.007256411015987396,
490 0.012649702839553356,
491 0.008141638711094856
492 ],
493 "mean": 0.010968165472149849,
494 "std": 0.0030406270010214335
495 },
496 {
497 "cfg": {
498 "lr": 0.01,
499 "epochs": 12,
500 "trap_lambda": 1.0
501 },
502 "per_seed": [
503 0.0021302360109984875,
504 0.003130449214950204,
505 0.003100306261330843,
506 0.0041888016276061535,
507 0.004465983249247074,
508 0.0038384413346648216,
509 0.0035596308298408985,
510 0.004667002707719803
511 ],
512 "mean": 0.0036351064045447856,
513 "std": 0.0008378638946201928
514 }
515 ],
516 "math_check": {
517 "formula_boundary": 1.2473835212243647,
518 "observed_grid_boundary": 1.25,
519 "absolute_error": 0.0026164787756353114,
520 "confirmed": true
521 },
522 "mechanism_signature": {
523 "prediction": "trap training should reduce one-step and long-horizon U violations",
524 "observed_one_step_violation": 0.257080078125,
525 "observed_100_step_violation": 0.0,
526 "observed_max_one_step_abs_coordinate": 0.9616694450378418,
527 "baseline_comparison_seed0": {
528 "mse": 0.0024895062670111656,
529 "one_step_violation": 0.8212890625,
530 "mean_violation": 0.9982080078125,
531 "max_norm": 3.4549334049224854,
532 "cloud_diameters": [
533 8.314221382141113,
534 3.868745750423841e-07,
535 2.440847026718984e-07
536 ]
537 },
538 "idea_comparison_seed0": {
539 "mse": 0.0020382918883115053,
540 "one_step_violation": 0.30322265625,
541 "mean_violation": 0.9889501953125,
542 "max_norm": 2.9793031215667725,
543 "cloud_diameters": [
544 8.197779655456543,
545 4.651096219276951e-07,
546 4.111343514523469e-07
547 ]
548 },
549 "confirmed": true
550 }
551 },
552 "bench_report": {
553 "baseline_sweep": [
554 {
555 "cfg": {
556 "lr": 0.001,
557 "epochs": 12,
558 "trap_lambda": 0.0
559 },
560 "per_seed": [
561 0.02792397141456604,
562 0.041552960872650146,
563 0.03038809262216091,
564 0.03133251145482063,
565 0.05597708746790886,
566 0.022259902209043503,
567 0.07735025882720947,
568 0.01893964223563671
569 ],
570 "mean": 0.038215553387999535,
571 "std": 0.019635395061371623
572 },
573 {
574 "cfg": {
575 "lr": 0.003,
576 "epochs": 12,
577 "trap_lambda": 0.0
578 },
579 "per_seed": [
580 0.009584075771272182,
581 0.013722526840865612,
582 0.008015970699489117,
583 0.011999157257378101,
584 0.01405302993953228,
585 0.00778926583006978,
586 0.013168145902454853,
587 0.007831467315554619
588 ],
589 "mean": 0.010770454944577068,
590 "std": 0.0027585833372694973
591 },
592 {
593 "cfg": {
594 "lr": 0.01,
595 "epochs": 12,
596 "trap_lambda": 0.0
597 },
598 "per_seed": [
599 0.0024895062670111656,
600 0.0038835888262838125,
601 0.003609295468777418,
602 0.003999150358140469,
603 0.0032447325065732002,
604 0.0038587753660976887,
605 0.008323567919433117,
606 0.0034713787026703358
607 ],
608 "mean": 0.004109999426873401,
609 "std": 0.0017693440291274862
610 },
611 {
612 "cfg": {
613 "lr": 0.001,
614 "epochs": 12,
615 "trap_lambda": 0.3
616 },
617 "per_seed": [
618 0.02792397141456604,
619 0.041552960872650146,
620 0.03038809262216091,
621 0.03133251145482063,
622 0.05597708746790886,
623 0.022259902209043503,
624 0.07735025882720947,
625 0.01893964223563671
626 ],
627 "mean": 0.038215553387999535,
628 "std": 0.019635395061371623
629 },
630 {
631 "cfg": {
632 "lr": 0.003,
633 "epochs": 12,
634 "trap_lambda": 0.3
635 },
636 "per_seed": [
637 0.009584075771272182,
638 0.013722526840865612,
639 0.008015970699489117,
640 0.011999157257378101,
641 0.01405302993953228,
642 0.00778926583006978,
643 0.013168145902454853,
644 0.007831467315554619
645 ],
646 "mean": 0.010770454944577068,
647 "std": 0.0027585833372694973
648 },
649 {
650 "cfg": {
651 "lr": 0.01,
652 "epochs": 12,
653 "trap_lambda": 0.3
654 },
655 "per_seed": [
656 0.0024895062670111656,
657 0.0038835888262838125,
658 0.003609295468777418,
659 0.003999150358140469,
660 0.0032447325065732002,
661 0.0038587753660976887,
662 0.008323567919433117,
663 0.0034713787026703358
664 ],
665 "mean": 0.004109999426873401,
666 "std": 0.0017693440291274862
667 },
668 {
669 "cfg": {
670 "lr": 0.001,
671 "epochs": 12,
672 "trap_lambda": 1.0
673 },
674 "per_seed": [
675 0.02792397141456604,
676 0.041552960872650146,
677 0.03038809262216091,
678 0.03133251145482063,
679 0.05597708746790886,
680 0.022259902209043503,
681 0.07735025882720947,
682 0.01893964223563671
683 ],
684 "mean": 0.038215553387999535,
685 "std": 0.019635395061371623
686 },
687 {
688 "cfg": {
689 "lr": 0.003,
690 "epochs": 12,
691 "trap_lambda": 1.0
692 },
693 "per_seed": [
694 0.009584075771272182,
695 0.013722526840865612,
696 0.008015970699489117,
697 0.011999157257378101,
698 0.01405302993953228,
699 0.00778926583006978,
700 0.013168145902454853,
701 0.007831467315554619
702 ],
703 "mean": 0.010770454944577068,
704 "std": 0.0027585833372694973
705 },
706 {
707 "cfg": {
708 "lr": 0.01,
709 "epochs": 12,
710 "trap_lambda": 1.0
711 },
712 "per_seed": [
713 0.0024895062670111656,
714 0.0038835888262838125,
715 0.003609295468777418,
716 0.003999150358140469,
717 0.0032447325065732002,
718 0.0038587753660976887,
719 0.008323567919433117,
720 0.0034713787026703358
721 ],
722 "mean": 0.004109999426873401,
723 "std": 0.0017693440291274862
724 }
725 ],
726 "idea_sweep": [
727 {
728 "cfg": {
729 "lr": 0.001,
730 "epochs": 12,
731 "trap_lambda": 0.0
732 },
733 "per_seed": [
734 0.02792397141456604,
735 0.041552960872650146,
736 0.03038809262216091,
737 0.03133251145482063,
738 0.05597708746790886,
739 0.022259902209043503,
740 0.07735025882720947,
741 0.01893964223563671
742 ],
743 "mean": 0.038215553387999535,
744 "std": 0.019635395061371623
745 },
746 {
747 "cfg": {
748 "lr": 0.003,
749 "epochs": 12,
750 "trap_lambda": 0.0
751 },
752 "per_seed": [
753 0.009584075771272182,
754 0.013722526840865612,
755 0.008015970699489117,
756 0.011999157257378101,
757 0.01405302993953228,
758 0.00778926583006978,
759 0.013168145902454853,
760 0.007831467315554619
761 ],
762 "mean": 0.010770454944577068,
763 "std": 0.0027585833372694973
764 },
765 {
766 "cfg": {
767 "lr": 0.01,
768 "epochs": 12,
769 "trap_lambda": 0.0
770 },
771 "per_seed": [
772 0.0024895062670111656,
773 0.0038835888262838125,
774 0.003609295468777418,
775 0.003999150358140469,
776 0.0032447325065732002,
777 0.0038587753660976887,
778 0.008323567919433117,
779 0.0034713787026703358
780 ],
781 "mean": 0.004109999426873401,
782 "std": 0.0017693440291274862
783 },
784 {
785 "cfg": {
786 "lr": 0.001,
787 "epochs": 12,
788 "trap_lambda": 0.3
789 },
790 "per_seed": [
791 0.029911765828728676,
792 0.043103478848934174,
793 0.029594196006655693,
794 0.03148665651679039,
795 0.056626711040735245,
796 0.023263107985258102,
797 0.07945620268583298,
798 0.019382713362574577
799 ],
800 "mean": 0.03910310403443873,
801 "std": 0.020086107344229626
802 },
803 {
804 "cfg": {
805 "lr": 0.003,
806 "epochs": 12,
807 "trap_lambda": 0.3
808 },
809 "per_seed": [
810 0.00927602220326662,
811 0.01410755980759859,
812 0.008517258800566196,
813 0.011783787980675697,
814 0.014786372892558575,
815 0.0075119370594620705,
816 0.012531742453575134,
817 0.008021011017262936
818 ],
819 "mean": 0.010816961526870728,
820 "std": 0.002850141431686086
821 },
822 {
823 "cfg": {
824 "lr": 0.01,
825 "epochs": 12,
826 "trap_lambda": 0.3
827 },
828 "per_seed": [
829 0.0020382918883115053,
830 0.002796109300106764,
831 0.00355388387106359,
832 0.0032515127677470446,
833 0.003949518781155348,
834 0.003303474746644497,
835 0.0037955872248858213,
836 0.004127324093133211
837 ],
838 "mean": 0.0033519628341309726,
839 "std": 0.0006806487150040169
840 },
841 {
842 "cfg": {
843 "lr": 0.001,
844 "epochs": 12,
845 "trap_lambda": 1.0
846 },
847 "per_seed": [
848 0.02979419380426407,
849 0.04312347620725632,
850 0.02988792397081852,
851 0.0319940410554409,
852 0.05702750012278557,
853 0.02334035187959671,
854 0.07955670356750488,
855 0.019355962052941322
856 ],
857 "mean": 0.039260019082576036,
858 "std": 0.020122028461056908
859 },
860 {
861 "cfg": {
862 "lr": 0.003,
863 "epochs": 12,
864 "trap_lambda": 1.0
865 },
866 "per_seed": [
867 0.009333031252026558,
868 0.014507703483104706,
869 0.00847854744642973,
870 0.012289805337786674,
871 0.015088483691215515,
872 0.007256411015987396,
873 0.012649702839553356,
874 0.008141638711094856
875 ],
876 "mean": 0.010968165472149849,
877 "std": 0.0030406270010214335
878 },
879 {
880 "cfg": {
881 "lr": 0.01,
882 "epochs": 12,
883 "trap_lambda": 1.0
884 },
885 "per_seed": [
886 0.0021302360109984875,
887 0.003130449214950204,
888 0.003100306261330843,
889 0.0041888016276061535,
890 0.004465983249247074,
891 0.0038384413346648216,
892 0.0035596308298408985,
893 0.004667002707719803
894 ],
895 "mean": 0.0036351064045447856,
896 "std": 0.0008378638946201928
897 }
898 ],
899 "paired_delta": -0.0007580365927424282,
900 "comparison_metric": "test_mse"
901 }
902}