Latent-Component Schrödinger Bridge / bench_report.json
Beats tuned baseline
1{
2 "bench_version": 1,
3 "track": "latent_mixture_transport",
4 "model": "mlp_tiny",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.01
10 },
11 "sweep": [
12 {
13 "per_seed": [
14 1.6584807634353638,
15 1.8023967742919922,
16 1.948538064956665,
17 1.7409632205963135,
18 1.6989234685897827,
19 2.0602176189422607,
20 1.787219524383545,
21 1.790303349494934
22 ],
23 "mean": 1.8108803480863571,
24 "std": 0.1239063765057777,
25 "config": {
26 "lr": 0.001
27 },
28 "signatures": [
29 {
30 "pred_mean": [
31 0.043234892189502716,
32 0.06797278672456741
33 ],
34 "pred_std": [
35 4.53477668762207,
36 1.0423473119735718
37 ],
38 "target_mean": [
39 0.10782764106988907,
40 -0.019209526479244232
41 ],
42 "target_std": [
43 4.536637783050537,
44 1.6189589500427246
45 ]
46 },
47 {
48 "pred_mean": [
49 0.21873944997787476,
50 -0.13643093407154083
51 ],
52 "pred_std": [
53 4.613937854766846,
54 1.0739085674285889
55 ],
56 "target_mean": [
57 0.28565511107444763,
58 -0.07340104132890701
59 ],
60 "target_std": [
61 4.524669170379639,
62 1.6303699016571045
63 ]
64 },
65 {
66 "pred_mean": [
67 0.24090030789375305,
68 -0.05906757339835167
69 ],
70 "pred_std": [
71 4.5369672775268555,
72 0.9227220416069031
73 ],
74 "target_mean": [
75 0.37497055530548096,
76 -0.06369480490684509
77 ],
78 "target_std": [
79 4.55037260055542,
80 1.7904703617095947
81 ]
82 },
83 {
84 "pred_mean": [
85 -0.006168227177113295,
86 -0.1031428799033165
87 ],
88 "pred_std": [
89 4.422000408172607,
90 0.9418965578079224
91 ],
92 "target_mean": [
93 0.032932933419942856,
94 -0.04044200852513313
95 ],
96 "target_std": [
97 4.487130165100098,
98 1.6359201669692993
99 ]
100 },
101 {
102 "pred_mean": [
103 0.41770225763320923,
104 -0.030527109280228615
105 ],
106 "pred_std": [
107 4.762247562408447,
108 1.0726734399795532
109 ],
110 "target_mean": [
111 0.3325880467891693,
112 0.018684253096580505
113 ],
114 "target_std": [
115 4.577158451080322,
116 1.662758231163025
117 ]
118 },
119 {
120 "pred_mean": [
121 0.3517788052558899,
122 0.025978371500968933
123 ],
124 "pred_std": [
125 4.633047103881836,
126 0.9402766823768616
127 ],
128 "target_mean": [
129 0.0659371018409729,
130 0.149443119764328
131 ],
132 "target_std": [
133 4.564814567565918,
134 1.66139554977417
135 ]
136 },
137 {
138 "pred_mean": [
139 0.08992262184619904,
140 -0.0018513768445700407
141 ],
142 "pred_std": [
143 4.472599506378174,
144 0.9126293063163757
145 ],
146 "target_mean": [
147 0.15029877424240112,
148 -0.006664793472737074
149 ],
150 "target_std": [
151 4.5857696533203125,
152 1.6789789199829102
153 ]
154 },
155 {
156 "pred_mean": [
157 -0.22141434252262115,
158 0.0017037141369655728
159 ],
160 "pred_std": [
161 4.363723278045654,
162 0.849623441696167
163 ],
164 "target_mean": [
165 -0.2404521107673645,
166 0.11181623488664627
167 ],
168 "target_std": [
169 4.359499454498291,
170 1.6317424774169922
171 ]
172 }
173 ]
174 },
175 {
176 "per_seed": [
177 1.4588760137557983,
178 1.3883919715881348,
179 1.5811883211135864,
180 1.4916973114013672,
181 1.4352669715881348,
182 1.693009614944458,
183 1.5048573017120361,
184 1.3684591054916382
185 ],
186 "mean": 1.4902183264493942,
187 "std": 0.09925487394239246,
188 "config": {
189 "lr": 0.003
190 },
191 "signatures": [
192 {
193 "pred_mean": [
194 0.06621045619249344,
195 -0.1021866425871849
196 ],
197 "pred_std": [
198 4.53108549118042,
199 1.266789436340332
200 ],
201 "target_mean": [
202 0.10782764106988907,
203 -0.019209526479244232
204 ],
205 "target_std": [
206 4.536637783050537,
207 1.6189589500427246
208 ]
209 },
210 {
211 "pred_mean": [
212 0.17134340107440948,
213 -0.04508264362812042
214 ],
215 "pred_std": [
216 4.447978973388672,
217 1.245232343673706
218 ],
219 "target_mean": [
220 0.28565511107444763,
221 -0.07340104132890701
222 ],
223 "target_std": [
224 4.524669170379639,
225 1.6303699016571045
226 ]
227 },
228 {
229 "pred_mean": [
230 0.30628085136413574,
231 -0.01341619249433279
232 ],
233 "pred_std": [
234 4.2159857749938965,
235 1.1691285371780396
236 ],
237 "target_mean": [
238 0.37497055530548096,
239 -0.06369480490684509
240 ],
241 "target_std": [
242 4.55037260055542,
243 1.7904703617095947
244 ]
245 },
246 {
247 "pred_mean": [
248 -0.03721538558602333,
249 -0.04432598128914833
250 ],
251 "pred_std": [
252 4.213113784790039,
253 1.017906904220581
254 ],
255 "target_mean": [
256 0.032932933419942856,
257 -0.04044200852513313
258 ],
259 "target_std": [
260 4.487130165100098,
261 1.6359201669692993
262 ]
263 },
264 {
265 "pred_mean": [
266 0.5395490527153015,
267 -0.13345450162887573
268 ],
269 "pred_std": [
270 4.522548675537109,
271 1.2905967235565186
272 ],
273 "target_mean": [
274 0.3325880467891693,
275 0.018684253096580505
276 ],
277 "target_std": [
278 4.577158451080322,
279 1.662758231163025
280 ]
281 },
282 {
283 "pred_mean": [
284 0.2691681385040283,
285 0.1501285880804062
286 ],
287 "pred_std": [
288 4.459775447845459,
289 1.1214773654937744
290 ],
291 "target_mean": [
292 0.0659371018409729,
293 0.149443119764328
294 ],
295 "target_std": [
296 4.564814567565918,
297 1.66139554977417
298 ]
299 },
300 {
301 "pred_mean": [
302 0.1690424382686615,
303 -0.08001205325126648
304 ],
305 "pred_std": [
306 4.409627914428711,
307 1.0867819786071777
308 ],
309 "target_mean": [
310 0.15029877424240112,
311 -0.006664793472737074
312 ],
313 "target_std": [
314 4.5857696533203125,
315 1.6789789199829102
316 ]
317 },
318 {
319 "pred_mean": [
320 -0.17967349290847778,
321 -0.042357780039310455
322 ],
323 "pred_std": [
324 4.187299728393555,
325 1.2296702861785889
326 ],
327 "target_mean": [
328 -0.2404521107673645,
329 0.11181623488664627
330 ],
331 "target_std": [
332 4.359499454498291,
333 1.6317424774169922
334 ]
335 }
336 ]
337 },
338 {
339 "per_seed": [
340 1.5505059957504272,
341 1.2503662109375,
342 1.5140531063079834,
343 1.4406028985977173,
344 1.48829984664917,
345 1.5539131164550781,
346 1.3900737762451172,
347 1.3093628883361816
348 ],
349 "mean": 1.4371472299098969,
350 "std": 0.1051543061191552,
351 "config": {
352 "lr": 0.01
353 },
354 "signatures": [
355 {
356 "pred_mean": [
357 0.2688068449497223,
358 -0.1351122260093689
359 ],
360 "pred_std": [
361 4.552779674530029,
362 1.5561680793762207
363 ],
364 "target_mean": [
365 0.10782764106988907,
366 -0.019209526479244232
367 ],
368 "target_std": [
369 4.536637783050537,
370 1.6189589500427246
371 ]
372 },
373 {
374 "pred_mean": [
375 0.24252046644687653,
376 0.0590505413711071
377 ],
378 "pred_std": [
379 4.520332336425781,
380 1.3489108085632324
381 ],
382 "target_mean": [
383 0.28565511107444763,
384 -0.07340104132890701
385 ],
386 "target_std": [
387 4.524669170379639,
388 1.6303699016571045
389 ]
390 },
391 {
392 "pred_mean": [
393 0.6274702548980713,
394 -0.20266346633434296
395 ],
396 "pred_std": [
397 4.23753023147583,
398 1.2169533967971802
399 ],
400 "target_mean": [
401 0.37497055530548096,
402 -0.06369480490684509
403 ],
404 "target_std": [
405 4.55037260055542,
406 1.7904703617095947
407 ]
408 },
409 {
410 "pred_mean": [
411 0.03630140423774719,
412 -0.12918049097061157
413 ],
414 "pred_std": [
415 4.289375305175781,
416 1.1092267036437988
417 ],
418 "target_mean": [
419 0.032932933419942856,
420 -0.04044200852513313
421 ],
422 "target_std": [
423 4.487130165100098,
424 1.6359201669692993
425 ]
426 },
427 {
428 "pred_mean": [
429 0.5265708565711975,
430 -0.3529617190361023
431 ],
432 "pred_std": [
433 4.702226161956787,
434 1.4276982545852661
435 ],
436 "target_mean": [
437 0.3325880467891693,
438 0.018684253096580505
439 ],
440 "target_std": [
441 4.577158451080322,
442 1.662758231163025
443 ]
444 },
445 {
446 "pred_mean": [
447 0.15553702414035797,
448 0.1446627378463745
449 ],
450 "pred_std": [
451 4.38116455078125,
452 1.2813775539398193
453 ],
454 "target_mean": [
455 0.0659371018409729,
456 0.149443119764328
457 ],
458 "target_std": [
459 4.564814567565918,
460 1.66139554977417
461 ]
462 },
463 {
464 "pred_mean": [
465 0.23208913207054138,
466 -0.04567822068929672
467 ],
468 "pred_std": [
469 4.284600734710693,
470 1.240485668182373
471 ],
472 "target_mean": [
473 0.15029877424240112,
474 -0.006664793472737074
475 ],
476 "target_std": [
477 4.5857696533203125,
478 1.6789789199829102
479 ]
480 },
481 {
482 "pred_mean": [
483 -0.31963932514190674,
484 -0.02888385020196438
485 ],
486 "pred_std": [
487 4.275836944580078,
488 1.2498278617858887
489 ],
490 "target_mean": [
491 -0.2404521107673645,
492 0.11181623488664627
493 ],
494 "target_std": [
495 4.359499454498291,
496 1.6317424774169922
497 ]
498 }
499 ]
500 }
501 ],
502 "full": {
503 "per_seed": [
504 1.5505059957504272,
505 1.2503662109375,
506 1.5140531063079834,
507 1.4406028985977173,
508 1.48829984664917,
509 1.5539131164550781,
510 1.3900737762451172,
511 1.3093628883361816
512 ],
513 "mean": 1.4371472299098969,
514 "std": 0.1051543061191552,
515 "config": {
516 "lr": 0.01
517 },
518 "signatures": [
519 {
520 "pred_mean": [
521 0.2688068449497223,
522 -0.1351122260093689
523 ],
524 "pred_std": [
525 4.552779674530029,
526 1.5561680793762207
527 ],
528 "target_mean": [
529 0.10782764106988907,
530 -0.019209526479244232
531 ],
532 "target_std": [
533 4.536637783050537,
534 1.6189589500427246
535 ]
536 },
537 {
538 "pred_mean": [
539 0.24252046644687653,
540 0.0590505413711071
541 ],
542 "pred_std": [
543 4.520332336425781,
544 1.3489108085632324
545 ],
546 "target_mean": [
547 0.28565511107444763,
548 -0.07340104132890701
549 ],
550 "target_std": [
551 4.524669170379639,
552 1.6303699016571045
553 ]
554 },
555 {
556 "pred_mean": [
557 0.6274702548980713,
558 -0.20266346633434296
559 ],
560 "pred_std": [
561 4.23753023147583,
562 1.2169533967971802
563 ],
564 "target_mean": [
565 0.37497055530548096,
566 -0.06369480490684509
567 ],
568 "target_std": [
569 4.55037260055542,
570 1.7904703617095947
571 ]
572 },
573 {
574 "pred_mean": [
575 0.03630140423774719,
576 -0.12918049097061157
577 ],
578 "pred_std": [
579 4.289375305175781,
580 1.1092267036437988
581 ],
582 "target_mean": [
583 0.032932933419942856,
584 -0.04044200852513313
585 ],
586 "target_std": [
587 4.487130165100098,
588 1.6359201669692993
589 ]
590 },
591 {
592 "pred_mean": [
593 0.5265708565711975,
594 -0.3529617190361023
595 ],
596 "pred_std": [
597 4.702226161956787,
598 1.4276982545852661
599 ],
600 "target_mean": [
601 0.3325880467891693,
602 0.018684253096580505
603 ],
604 "target_std": [
605 4.577158451080322,
606 1.662758231163025
607 ]
608 },
609 {
610 "pred_mean": [
611 0.15553702414035797,
612 0.1446627378463745
613 ],
614 "pred_std": [
615 4.38116455078125,
616 1.2813775539398193
617 ],
618 "target_mean": [
619 0.0659371018409729,
620 0.149443119764328
621 ],
622 "target_std": [
623 4.564814567565918,
624 1.66139554977417
625 ]
626 },
627 {
628 "pred_mean": [
629 0.23208913207054138,
630 -0.04567822068929672
631 ],
632 "pred_std": [
633 4.284600734710693,
634 1.240485668182373
635 ],
636 "target_mean": [
637 0.15029877424240112,
638 -0.006664793472737074
639 ],
640 "target_std": [
641 4.5857696533203125,
642 1.6789789199829102
643 ]
644 },
645 {
646 "pred_mean": [
647 -0.31963932514190674,
648 -0.02888385020196438
649 ],
650 "pred_std": [
651 4.275836944580078,
652 1.2498278617858887
653 ],
654 "target_mean": [
655 -0.2404521107673645,
656 0.11181623488664627
657 ],
658 "target_std": [
659 4.359499454498291,
660 1.6317424774169922
661 ]
662 }
663 ]
664 }
665 },
666 "idea": {
667 "per_seed": [
668 0.1718381643295288,
669 0.13643871247768402,
670 0.17405973374843597,
671 0.17571242153644562,
672 0.14317259192466736,
673 0.19144359230995178,
674 0.16002926230430603,
675 0.1587807536125183
676 ],
677 "mean": 0.16393440403044224,
678 "std": 0.01691650453456634,
679 "config": {
680 "lr": 0.01,
681 "epsilon": 0.1,
682 "tau": 3.0
683 },
684 "signatures": [
685 {
686 "pred_mean": [
687 0.06780313700437546,
688 -0.053052350878715515
689 ],
690 "pred_std": [
691 2.044187545776367,
692 1.0252753496170044
693 ],
694 "target_mean": [
695 0.04435388743877411,
696 -0.012574340216815472
697 ],
698 "target_std": [
699 2.0263524055480957,
700 1.0520573854446411
701 ]
702 },
703 {
704 "pred_mean": [
705 0.09832692891359329,
706 0.015258383937180042
707 ],
708 "pred_std": [
709 2.0508484840393066,
710 1.0656163692474365
711 ],
712 "target_mean": [
713 0.13986718654632568,
714 -0.07649480551481247
715 ],
716 "target_std": [
717 2.022428512573242,
718 1.0598915815353394
719 ]
720 },
721 {
722 "pred_mean": [
723 0.17367428541183472,
724 -0.08788446336984634
725 ],
726 "pred_std": [
727 1.9044790267944336,
728 0.9941457509994507
729 ],
730 "target_mean": [
731 0.12951253354549408,
732 -0.05628962442278862
733 ],
734 "target_std": [
735 2.026750326156616,
736 1.075918436050415
737 ]
738 },
739 {
740 "pred_mean": [
741 0.0790666937828064,
742 -0.0549687035381794
743 ],
744 "pred_std": [
745 1.9015724658966064,
746 0.9414774775505066
747 ],
748 "target_mean": [
749 0.028750987723469734,
750 -0.007199820131063461
751 ],
752 "target_std": [
753 2.000044822692871,
754 1.0400867462158203
755 ]
756 },
757 {
758 "pred_mean": [
759 0.1755228042602539,
760 -0.08683786541223526
761 ],
762 "pred_std": [
763 2.0293917655944824,
764 1.0066801309585571
765 ],
766 "target_mean": [
767 0.13575679063796997,
768 -0.049245692789554596
769 ],
770 "target_std": [
771 2.0310559272766113,
772 1.0684142112731934
773 ]
774 },
775 {
776 "pred_mean": [
777 0.0953637957572937,
778 0.01983083039522171
779 ],
780 "pred_std": [
781 1.9573743343353271,
782 0.9901137948036194
783 ],
784 "target_mean": [
785 0.02048776112496853,
786 0.015579056926071644
787 ],
788 "target_std": [
789 2.0075297355651855,
790 1.0683743953704834
791 ]
792 },
793 {
794 "pred_mean": [
795 0.009151271544396877,
796 -0.024620816111564636
797 ],
798 "pred_std": [
799 1.9104688167572021,
800 0.9687676429748535
801 ],
802 "target_mean": [
803 0.07076805084943771,
804 -0.020372634753584862
805 ],
806 "target_std": [
807 2.042053699493408,
808 1.0628199577331543
809 ]
810 },
811 {
812 "pred_mean": [
813 -0.15449641644954681,
814 0.07149290293455124
815 ],
816 "pred_std": [
817 1.8906241655349731,
818 0.9696427583694458
819 ],
820 "target_mean": [
821 -0.13105426728725433,
822 0.08039151132106781
823 ],
824 "target_std": [
825 1.9721097946166992,
826 1.02797532081604
827 ]
828 }
829 ]
830 },
831 "comparison": {
832 "delta_mean": -1.2732128258794546,
833 "idea_wins": 8,
834 "n_pairs": 8,
835 "per_seed_diffs": [
836 -1.3786678314208984,
837 -1.113927498459816,
838 -1.3399933725595474,
839 -1.2648904770612717,
840 -1.3451272547245026,
841 -1.3624695241451263,
842 -1.2300445139408112,
843 -1.1505821347236633
844 ],
845 "p_value": 0.0081,
846 "mde": 0.08424709934543327,
847 "mde_rel_pct": 5.862106372408011,
848 "verdict": "idea better (significant)",
849 "system_worked": true
850 },
851 "mechanism_signature": {
852 "track_structure": "endpoint Gaussian-mixture transport",
853 "signature": {
854 "prediction": "component smoothing should reduce target-label assignment variance",
855 "predicted": 0.07482332631479949,
856 "observed": 0.06172130233608186,
857 "confirmed": false
858 },
859 "custom_track": {
860 "name": "latent_mixture_transport",
861 "file": "custom_track.py",
862 "domain": "diffusion-sampling"
863 }
864 }
865}