wuff-mann commited on
Commit
4f29f4e
·
verified ·
1 Parent(s): 4dad2f8

CIDM-v2 A2 report.json

Browse files
Files changed (1) hide show
  1. research/v2/A2/report.json +952 -0
research/v2/A2/report.json ADDED
@@ -0,0 +1,952 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "A2_NOT_QUALIFIED_NEEDS_REVIEW",
3
+ "investment": "PRO6000_BLOCKED_SYNTHETIC_ONLY",
4
+ "config": {
5
+ "workdir": "/content/CIDM_v2_A2",
6
+ "epochs": 10,
7
+ "steps_per_epoch": 48,
8
+ "batch": 4,
9
+ "H": 24,
10
+ "W": 32,
11
+ "d": 24,
12
+ "hidden": 32,
13
+ "lr": 0.0004,
14
+ "seed": 20261010,
15
+ "cf_weight": 0.55
16
+ },
17
+ "params": 41356,
18
+ "device": "cuda",
19
+ "best_epoch": 9,
20
+ "gate": {
21
+ "id_field_skill": true,
22
+ "ood_12_field_skill": true,
23
+ "ood_24_field_skill": true,
24
+ "ood_12_responds_to_matched_counterfactual": false,
25
+ "ood_velocity_alone_causes_predictive_response": false,
26
+ "ood_source_alone_causes_predictive_response": true,
27
+ "ood_24_correct_controls_matter": true,
28
+ "ood_36_has_nonnegative_counterfactual_skill": true,
29
+ "synthetic_port_neutral_and_pair_transfer_closed": true
30
+ },
31
+ "results": {
32
+ "id12": {
33
+ "mse": 0.011008506175130606,
34
+ "persistence": 0.02467968373093754,
35
+ "swapped_controls": 0.056886752136051655,
36
+ "no_soc": 0.01399493106873706,
37
+ "no_sync": 0.010854822874534875,
38
+ "no_flux": 0.01162089838180691,
39
+ "zero_source": 0.02424306597094983,
40
+ "delta_mse": 0.035967894131317735,
41
+ "delta_truth_power": 0.08876745495945215,
42
+ "delta_pred_power": 0.038956932723522186,
43
+ "rps": 0.10431083757430315,
44
+ "uniform_rps": 0.18210653215646744,
45
+ "budget": 4.842877388000488e-08,
46
+ "latent_rms": 0.535042367875576,
47
+ "n_seeds": 8,
48
+ "relative_mse_to_persistence": 0.44605539905403035,
49
+ "counterfactual_skill": 0.5948076448992774,
50
+ "swapped_control_penalty": 4.1675269315436205,
51
+ "no_soc_penalty": 0.2712833917787236,
52
+ "no_sync_penalty": -0.013960413715615405,
53
+ "no_flux_penalty": 0.0556290015133718,
54
+ "no_source_penalty": 1.2022121426172707,
55
+ "rps_vs_uniform": 0.5728011858722257,
56
+ "rows": [
57
+ {
58
+ "seed": 910000,
59
+ "mse": 0.009263951797038317,
60
+ "persistence": 0.022598392330110073,
61
+ "swapped_controls": 0.05502618849277496,
62
+ "no_soc": 0.012785566970705986,
63
+ "no_sync": 0.008873223327100277,
64
+ "no_flux": 0.009751887992024422,
65
+ "zero_source": 0.022588597610592842,
66
+ "delta_mse": 0.02986116148531437,
67
+ "delta_truth_power": 0.08342001587152481,
68
+ "delta_pred_power": 0.03796563297510147,
69
+ "rps": 0.10140340030193329,
70
+ "uniform_rps": 0.1883748471736908,
71
+ "budget": 5.960464477539063e-08,
72
+ "latent_rms": 0.5425789952278137
73
+ },
74
+ {
75
+ "seed": 910001,
76
+ "mse": 0.009976076893508434,
77
+ "persistence": 0.023790083825588226,
78
+ "swapped_controls": 0.05671653151512146,
79
+ "no_soc": 0.01298708375543356,
80
+ "no_sync": 0.009749122429639101,
81
+ "no_flux": 0.010421553160995245,
82
+ "zero_source": 0.02353616151958704,
83
+ "delta_mse": 0.03329237177968025,
84
+ "delta_truth_power": 0.08755321800708771,
85
+ "delta_pred_power": 0.03922006115317345,
86
+ "rps": 0.10330619663000107,
87
+ "uniform_rps": 0.18195394426584244,
88
+ "budget": 2.9802322387695312e-08,
89
+ "latent_rms": 0.5344235897064209
90
+ },
91
+ {
92
+ "seed": 910002,
93
+ "mse": 0.0076738749630749226,
94
+ "persistence": 0.020854060538113117,
95
+ "swapped_controls": 0.05258640646934509,
96
+ "no_soc": 0.010751384310424328,
97
+ "no_sync": 0.007407581899315119,
98
+ "no_flux": 0.00804087333381176,
99
+ "zero_source": 0.020705181173980236,
100
+ "delta_mse": 0.025028085336089134,
101
+ "delta_truth_power": 0.07741687446832657,
102
+ "delta_pred_power": 0.03743627667427063,
103
+ "rps": 0.10251637548208237,
104
+ "uniform_rps": 0.18363037705421448,
105
+ "budget": 2.9802322387695312e-08,
106
+ "latent_rms": 0.5432775616645813
107
+ },
108
+ {
109
+ "seed": 910003,
110
+ "mse": 0.016451374161988497,
111
+ "persistence": 0.03190521337091923,
112
+ "swapped_controls": 0.06669662520289421,
113
+ "no_soc": 0.019024976529181004,
114
+ "no_sync": 0.016599673312157393,
115
+ "no_flux": 0.017371049150824547,
116
+ "zero_source": 0.030769897624850273,
117
+ "delta_mse": 0.0542491190135479,
118
+ "delta_truth_power": 0.11039359122514725,
119
+ "delta_pred_power": 0.044346027076244354,
120
+ "rps": 0.10890697687864304,
121
+ "uniform_rps": 0.17903240025043488,
122
+ "budget": 5.960464477539063e-08,
123
+ "latent_rms": 0.5261033177375793
124
+ },
125
+ {
126
+ "seed": 910004,
127
+ "mse": 0.011525016278028488,
128
+ "persistence": 0.023892446421086788,
129
+ "swapped_controls": 0.05168454721570015,
130
+ "no_soc": 0.014108978677541018,
131
+ "no_sync": 0.011574558913707733,
132
+ "no_flux": 0.012295011896640062,
133
+ "zero_source": 0.023068313486874104,
134
+ "delta_mse": 0.03663868084549904,
135
+ "delta_truth_power": 0.082478366792202,
136
+ "delta_pred_power": 0.03447937220335007,
137
+ "rps": 0.10434799641370773,
138
+ "uniform_rps": 0.17735596001148224,
139
+ "budget": 5.960464477539063e-08,
140
+ "latent_rms": 0.5318727493286133
141
+ },
142
+ {
143
+ "seed": 910005,
144
+ "mse": 0.00911463238298893,
145
+ "persistence": 0.022160018794238567,
146
+ "swapped_controls": 0.05407794378697872,
147
+ "no_soc": 0.012031451798975468,
148
+ "no_sync": 0.008972926530987024,
149
+ "no_flux": 0.009875132702291012,
150
+ "zero_source": 0.0219778036698699,
151
+ "delta_mse": 0.028947284445166588,
152
+ "delta_truth_power": 0.08033662289381027,
153
+ "delta_pred_power": 0.03853728994727135,
154
+ "rps": 0.105803482234478,
155
+ "uniform_rps": 0.17623697966337204,
156
+ "budget": 5.960464477539063e-08,
157
+ "latent_rms": 0.5394272208213806
158
+ },
159
+ {
160
+ "seed": 910006,
161
+ "mse": 0.011921252589672804,
162
+ "persistence": 0.028491022065281868,
163
+ "swapped_controls": 0.0678306482732296,
164
+ "no_soc": 0.01499450160190463,
165
+ "no_sync": 0.011671120300889015,
166
+ "no_flux": 0.01249442994594574,
167
+ "zero_source": 0.028092448599636555,
168
+ "delta_mse": 0.03958352282643318,
169
+ "delta_truth_power": 0.10446459800004959,
170
+ "delta_pred_power": 0.046937715262174606,
171
+ "rps": 0.10634662210941315,
172
+ "uniform_rps": 0.18871663510799408,
173
+ "budget": 2.9802322387695312e-08,
174
+ "latent_rms": 0.5329263806343079
175
+ },
176
+ {
177
+ "seed": 910007,
178
+ "mse": 0.012141870334744453,
179
+ "persistence": 0.023746232502162457,
180
+ "swapped_controls": 0.05047512613236904,
181
+ "no_soc": 0.015275504905730486,
182
+ "no_sync": 0.01199037628248334,
183
+ "no_flux": 0.012717248871922493,
184
+ "zero_source": 0.02320612408220768,
185
+ "delta_mse": 0.04014292731881142,
186
+ "delta_truth_power": 0.08407635241746902,
187
+ "delta_pred_power": 0.03273308649659157,
188
+ "rps": 0.10185565054416656,
189
+ "uniform_rps": 0.18155111372470856,
190
+ "budget": 5.960464477539063e-08,
191
+ "latent_rms": 0.5297291278839111
192
+ }
193
+ ]
194
+ },
195
+ "ood12": {
196
+ "mse": 0.039892679546028376,
197
+ "persistence": 0.05467459582723677,
198
+ "swapped_controls": 0.0772428005002439,
199
+ "no_soc": 0.04061369597911835,
200
+ "no_sync": 0.04125459771603346,
201
+ "no_flux": 0.042957893339917064,
202
+ "zero_source": 0.049780311062932014,
203
+ "delta_mse": 0.0971931628882885,
204
+ "delta_truth_power": 0.13594767916947603,
205
+ "delta_pred_power": 0.03594572586007416,
206
+ "rps": 0.11543341539800167,
207
+ "uniform_rps": 0.17506663501262665,
208
+ "budget": 5.029141902923584e-08,
209
+ "latent_rms": 0.5131536722183228,
210
+ "n_seeds": 8,
211
+ "relative_mse_to_persistence": 0.7296383071963265,
212
+ "counterfactual_skill": 0.2850693481341091,
213
+ "swapped_control_penalty": 0.9362650335663907,
214
+ "no_soc_penalty": 0.018073903314969275,
215
+ "no_sync_penalty": 0.03413955105306221,
216
+ "no_flux_penalty": 0.07683649804350767,
217
+ "no_source_penalty": 0.24785578781428397,
218
+ "rps_vs_uniform": 0.6593684478465931,
219
+ "rows": [
220
+ {
221
+ "seed": 990000,
222
+ "mse": 0.04332285188138485,
223
+ "persistence": 0.057778697460889816,
224
+ "swapped_controls": 0.07864677906036377,
225
+ "no_soc": 0.04401409439742565,
226
+ "no_sync": 0.04475691728293896,
227
+ "no_flux": 0.046247437596321106,
228
+ "zero_source": 0.05263460427522659,
229
+ "delta_mse": 0.10285541415214539,
230
+ "delta_truth_power": 0.13825184106826782,
231
+ "delta_pred_power": 0.0352514311671257,
232
+ "rps": 0.11596968024969101,
233
+ "uniform_rps": 0.17137858271598816,
234
+ "budget": 5.960464477539063e-08,
235
+ "latent_rms": 0.5038571357727051
236
+ },
237
+ {
238
+ "seed": 990001,
239
+ "mse": 0.04576772637665272,
240
+ "persistence": 0.059268439188599586,
241
+ "swapped_controls": 0.07857472077012062,
242
+ "no_soc": 0.04648067057132721,
243
+ "no_sync": 0.04712873883545399,
244
+ "no_flux": 0.04855967313051224,
245
+ "zero_source": 0.05444709397852421,
246
+ "delta_mse": 0.11047650128602982,
247
+ "delta_truth_power": 0.14540700614452362,
248
+ "delta_pred_power": 0.030683478340506554,
249
+ "rps": 0.11210902780294418,
250
+ "uniform_rps": 0.17637939751148224,
251
+ "budget": 5.960464477539063e-08,
252
+ "latent_rms": 0.5141065120697021
253
+ },
254
+ {
255
+ "seed": 990002,
256
+ "mse": 0.03770630620419979,
257
+ "persistence": 0.05231323465704918,
258
+ "swapped_controls": 0.07728702947497368,
259
+ "no_soc": 0.038595184683799744,
260
+ "no_sync": 0.038792211562395096,
261
+ "no_flux": 0.040590425953269005,
262
+ "zero_source": 0.04821939393877983,
263
+ "delta_mse": 0.09741270542144775,
264
+ "delta_truth_power": 0.14083218574523926,
265
+ "delta_pred_power": 0.035741981118917465,
266
+ "rps": 0.11347026005387306,
267
+ "uniform_rps": 0.1776529997587204,
268
+ "budget": 5.960464477539063e-08,
269
+ "latent_rms": 0.5114507675170898
270
+ },
271
+ {
272
+ "seed": 990003,
273
+ "mse": 0.03233289532363415,
274
+ "persistence": 0.047578928992152214,
275
+ "swapped_controls": 0.07223541662096977,
276
+ "no_soc": 0.03292056545615196,
277
+ "no_sync": 0.03367650881409645,
278
+ "no_flux": 0.03564311936497688,
279
+ "zero_source": 0.04296598397195339,
280
+ "delta_mse": 0.07846281677484512,
281
+ "delta_truth_power": 0.11985381692647934,
282
+ "delta_pred_power": 0.03841404989361763,
283
+ "rps": 0.11751919984817505,
284
+ "uniform_rps": 0.17224528640508652,
285
+ "budget": 4.470348358154297e-08,
286
+ "latent_rms": 0.5184615850448608
287
+ },
288
+ {
289
+ "seed": 990004,
290
+ "mse": 0.04005730152130127,
291
+ "persistence": 0.05557638593018055,
292
+ "swapped_controls": 0.07896332070231438,
293
+ "no_soc": 0.04074270837008953,
294
+ "no_sync": 0.04146366007626057,
295
+ "no_flux": 0.04324926622211933,
296
+ "zero_source": 0.05029783770442009,
297
+ "delta_mse": 0.097772017121315,
298
+ "delta_truth_power": 0.13843822479248047,
299
+ "delta_pred_power": 0.03714582696557045,
300
+ "rps": 0.11560801416635513,
301
+ "uniform_rps": 0.1802571639418602,
302
+ "budget": 5.960464477539063e-08,
303
+ "latent_rms": 0.516528844833374
304
+ },
305
+ {
306
+ "seed": 990005,
307
+ "mse": 0.033339599147439,
308
+ "persistence": 0.047494012862443924,
309
+ "swapped_controls": 0.07030823081731796,
310
+ "no_soc": 0.034209366887807846,
311
+ "no_sync": 0.034523213282227516,
312
+ "no_flux": 0.035924216732382774,
313
+ "zero_source": 0.04324863292276859,
314
+ "delta_mse": 0.08637804538011551,
315
+ "delta_truth_power": 0.1250794678926468,
316
+ "delta_pred_power": 0.03523583710193634,
317
+ "rps": 0.11327286064624786,
318
+ "uniform_rps": 0.17197266221046448,
319
+ "budget": 4.470348358154297e-08,
320
+ "latent_rms": 0.5153510570526123
321
+ },
322
+ {
323
+ "seed": 990006,
324
+ "mse": 0.04367976449429989,
325
+ "persistence": 0.058086466044187546,
326
+ "swapped_controls": 0.07801331952214241,
327
+ "no_soc": 0.044347064569592476,
328
+ "no_sync": 0.04522436670958996,
329
+ "no_flux": 0.04700975678861141,
330
+ "zero_source": 0.052610255777835846,
331
+ "delta_mse": 0.10273427516222,
332
+ "delta_truth_power": 0.1367664486169815,
333
+ "delta_pred_power": 0.03463492915034294,
334
+ "rps": 0.1166013777256012,
335
+ "uniform_rps": 0.1751709058880806,
336
+ "budget": 4.470348358154297e-08,
337
+ "latent_rms": 0.51296067237854
338
+ },
339
+ {
340
+ "seed": 990007,
341
+ "mse": 0.04293499141931534,
342
+ "persistence": 0.05930060148239136,
343
+ "swapped_controls": 0.08391358703374863,
344
+ "no_soc": 0.04359991289675236,
345
+ "no_sync": 0.04447116516530514,
346
+ "no_flux": 0.04643925093114376,
347
+ "zero_source": 0.05381868593394756,
348
+ "delta_mse": 0.10145352780818939,
349
+ "delta_truth_power": 0.14295244216918945,
350
+ "delta_pred_power": 0.04045827314257622,
351
+ "rps": 0.11891690269112587,
352
+ "uniform_rps": 0.1754760816693306,
353
+ "budget": 2.9802322387695312e-08,
354
+ "latent_rms": 0.5125128030776978
355
+ }
356
+ ]
357
+ },
358
+ "ood12_velocity": {
359
+ "mse": 0.04031829256564379,
360
+ "persistence": 0.05500972387380898,
361
+ "swapped_controls": 0.045304276747629046,
362
+ "no_soc": 0.041045802645385265,
363
+ "no_sync": 0.04175343387760222,
364
+ "no_flux": 0.04352927068248391,
365
+ "zero_source": 0.04958223388530314,
366
+ "delta_mse": 0.09702479653060436,
367
+ "delta_truth_power": 0.10565043333917856,
368
+ "delta_pred_power": 0.0013463262293953449,
369
+ "rps": 0.11478224163874984,
370
+ "uniform_rps": 0.17541351448744535,
371
+ "budget": 5.029141902923584e-08,
372
+ "latent_rms": 0.5131536722183228,
373
+ "n_seeds": 8,
374
+ "relative_mse_to_persistence": 0.7329302844372207,
375
+ "counterfactual_skill": 0.08164317491138517,
376
+ "swapped_control_penalty": 0.12366555884943242,
377
+ "no_soc_penalty": 0.018044168873396327,
378
+ "no_sync_penalty": 0.03559528989532046,
379
+ "no_flux_penalty": 0.07964072664069799,
380
+ "no_source_penalty": 0.22977017949300227,
381
+ "rps_vs_uniform": 0.6543523284060591,
382
+ "rows": [
383
+ {
384
+ "seed": 990000,
385
+ "mse": 0.043886398896574974,
386
+ "persistence": 0.058156199753284454,
387
+ "swapped_controls": 0.048664914444088936,
388
+ "no_soc": 0.04460535757243633,
389
+ "no_sync": 0.04538791812956333,
390
+ "no_flux": 0.04696759581565857,
391
+ "zero_source": 0.05253053084015846,
392
+ "delta_mse": 0.1031222939491272,
393
+ "delta_truth_power": 0.11126392334699631,
394
+ "delta_pred_power": 0.0014154003001749516,
395
+ "rps": 0.11513421684503555,
396
+ "uniform_rps": 0.1719360277056694,
397
+ "budget": 5.960464477539063e-08,
398
+ "latent_rms": 0.5038571357727051
399
+ },
400
+ {
401
+ "seed": 990001,
402
+ "mse": 0.04602149687707424,
403
+ "persistence": 0.05968763865530491,
404
+ "swapped_controls": 0.05054101720452309,
405
+ "no_soc": 0.046709608286619186,
406
+ "no_sync": 0.04742489568889141,
407
+ "no_flux": 0.048900311812758446,
408
+ "zero_source": 0.05450100265443325,
409
+ "delta_mse": 0.10940007865428925,
410
+ "delta_truth_power": 0.11727076023817062,
411
+ "delta_pred_power": 0.0011683464981615543,
412
+ "rps": 0.11230830103158951,
413
+ "uniform_rps": 0.17688395828008652,
414
+ "budget": 5.960464477539063e-08,
415
+ "latent_rms": 0.5141065120697021
416
+ },
417
+ {
418
+ "seed": 990002,
419
+ "mse": 0.03829500824213028,
420
+ "persistence": 0.052762093022465706,
421
+ "swapped_controls": 0.04313880205154419,
422
+ "no_soc": 0.03925335220992565,
423
+ "no_sync": 0.0394863486289978,
424
+ "no_flux": 0.04131999798119068,
425
+ "zero_source": 0.04800746217370033,
426
+ "delta_mse": 0.09648338705301285,
427
+ "delta_truth_power": 0.10501202940940857,
428
+ "delta_pred_power": 0.0011589484056457877,
429
+ "rps": 0.11225800216197968,
430
+ "uniform_rps": 0.17810872197151184,
431
+ "budget": 5.960464477539063e-08,
432
+ "latent_rms": 0.5114507675170898
433
+ },
434
+ {
435
+ "seed": 990003,
436
+ "mse": 0.033004505559802055,
437
+ "persistence": 0.04799623414874077,
438
+ "swapped_controls": 0.0384201854467392,
439
+ "no_soc": 0.03362441807985306,
440
+ "no_sync": 0.03444576822221279,
441
+ "no_flux": 0.036548325791954994,
442
+ "zero_source": 0.042718105018138885,
443
+ "delta_mse": 0.07980544120073318,
444
+ "delta_truth_power": 0.0891885906457901,
445
+ "delta_pred_power": 0.0014482077676802874,
446
+ "rps": 0.11651470512151718,
447
+ "uniform_rps": 0.17253011465072632,
448
+ "budget": 4.470348358154297e-08,
449
+ "latent_rms": 0.5184615850448608
450
+ },
451
+ {
452
+ "seed": 990004,
453
+ "mse": 0.04135022684931755,
454
+ "persistence": 0.05643124133348465,
455
+ "swapped_controls": 0.04674007371068001,
456
+ "no_soc": 0.04216520115733147,
457
+ "no_sync": 0.042855849489569664,
458
+ "no_flux": 0.04477420635521412,
459
+ "zero_source": 0.05071256123483181,
460
+ "delta_mse": 0.10004458576440811,
461
+ "delta_truth_power": 0.1094079464673996,
462
+ "delta_pred_power": 0.0014163164887577295,
463
+ "rps": 0.11518213897943497,
464
+ "uniform_rps": 0.18057454377412796,
465
+ "budget": 5.960464477539063e-08,
466
+ "latent_rms": 0.516528844833374
467
+ },
468
+ {
469
+ "seed": 990005,
470
+ "mse": 0.03366255760192871,
471
+ "persistence": 0.04791481792926788,
472
+ "swapped_controls": 0.037964774295687675,
473
+ "no_soc": 0.034480029717087746,
474
+ "no_sync": 0.034896668046712875,
475
+ "no_flux": 0.03636929951608181,
476
+ "zero_source": 0.04317535273730755,
477
+ "delta_mse": 0.08654064685106277,
478
+ "delta_truth_power": 0.09396582096815109,
479
+ "delta_pred_power": 0.0011792422737926245,
480
+ "rps": 0.11275765299797058,
481
+ "uniform_rps": 0.17222494632005692,
482
+ "budget": 4.470348358154297e-08,
483
+ "latent_rms": 0.5153510570526123
484
+ },
485
+ {
486
+ "seed": 990006,
487
+ "mse": 0.04339301027357578,
488
+ "persistence": 0.05800280719995499,
489
+ "swapped_controls": 0.04849734716117382,
490
+ "no_soc": 0.04398982413113117,
491
+ "no_sync": 0.04499006271362305,
492
+ "no_flux": 0.046842243522405624,
493
+ "zero_source": 0.05179947800934315,
494
+ "delta_mse": 0.10117708891630173,
495
+ "delta_truth_power": 0.1099454015493393,
496
+ "delta_pred_power": 0.0014403666136786342,
497
+ "rps": 0.11554206162691116,
498
+ "uniform_rps": 0.17512207478284836,
499
+ "budget": 4.470348358154297e-08,
500
+ "latent_rms": 0.51296067237854
501
+ },
502
+ {
503
+ "seed": 990007,
504
+ "mse": 0.042933136224746704,
505
+ "persistence": 0.05912675894796848,
506
+ "swapped_controls": 0.04846709966659546,
507
+ "no_soc": 0.04353863000869751,
508
+ "no_sync": 0.044539960101246834,
509
+ "no_flux": 0.04651218466460705,
510
+ "zero_source": 0.05321337841451168,
511
+ "delta_mse": 0.09962484985589981,
512
+ "delta_truth_power": 0.10914899408817291,
513
+ "delta_pred_power": 0.0015437814872711897,
514
+ "rps": 0.11856085434556007,
515
+ "uniform_rps": 0.17592772841453552,
516
+ "budget": 2.9802322387695312e-08,
517
+ "latent_rms": 0.5125128030776978
518
+ }
519
+ ]
520
+ },
521
+ "ood12_source": {
522
+ "mse": 0.03984345868229866,
523
+ "persistence": 0.05438381736166775,
524
+ "swapped_controls": 0.076820298563689,
525
+ "no_soc": 0.04049747111275792,
526
+ "no_sync": 0.041196447564288974,
527
+ "no_flux": 0.04286435921676457,
528
+ "zero_source": 0.0494594054762274,
529
+ "delta_mse": 0.012768498971126974,
530
+ "delta_truth_power": 0.05101336445659399,
531
+ "delta_pred_power": 0.03570881602354348,
532
+ "rps": 0.11470822663977742,
533
+ "uniform_rps": 0.17427826207131147,
534
+ "budget": 5.029141902923584e-08,
535
+ "latent_rms": 0.5131536722183228,
536
+ "n_seeds": 8,
537
+ "relative_mse_to_persistence": 0.7326344603088158,
538
+ "counterfactual_skill": 0.7497028649817564,
539
+ "swapped_control_penalty": 0.9280529628773948,
540
+ "no_soc_penalty": 0.016414549642243204,
541
+ "no_sync_penalty": 0.03395761629979699,
542
+ "no_flux_penalty": 0.07581923443327998,
543
+ "no_source_penalty": 0.24134317431134167,
544
+ "rps_vs_uniform": 0.6581900994218138,
545
+ "rows": [
546
+ {
547
+ "seed": 990000,
548
+ "mse": 0.04279658384621143,
549
+ "persistence": 0.056998154148459435,
550
+ "swapped_controls": 0.07762490212917328,
551
+ "no_soc": 0.04315839149057865,
552
+ "no_sync": 0.04423980601131916,
553
+ "no_flux": 0.045617448166012764,
554
+ "zero_source": 0.051600709557533264,
555
+ "delta_mse": 0.011469184421002865,
556
+ "delta_truth_power": 0.04659188538789749,
557
+ "delta_pred_power": 0.03453393653035164,
558
+ "rps": 0.11546342819929123,
559
+ "uniform_rps": 0.17060546576976776,
560
+ "budget": 5.960464477539063e-08,
561
+ "latent_rms": 0.5038571357727051
562
+ },
563
+ {
564
+ "seed": 990001,
565
+ "mse": 0.04561903700232506,
566
+ "persistence": 0.05943330004811287,
567
+ "swapped_controls": 0.07834517955780029,
568
+ "no_soc": 0.04610414244234562,
569
+ "no_sync": 0.046937260776758194,
570
+ "no_flux": 0.04841705597937107,
571
+ "zero_source": 0.05436384677886963,
572
+ "delta_mse": 0.012035636231303215,
573
+ "delta_truth_power": 0.046587664633989334,
574
+ "delta_pred_power": 0.030900252982974052,
575
+ "rps": 0.11367041245102882,
576
+ "uniform_rps": 0.17627359926700592,
577
+ "budget": 5.960464477539063e-08,
578
+ "latent_rms": 0.5141065120697021
579
+ },
580
+ {
581
+ "seed": 990002,
582
+ "mse": 0.037735434249043465,
583
+ "persistence": 0.051150163635611534,
584
+ "swapped_controls": 0.07388226687908173,
585
+ "no_soc": 0.03850729390978813,
586
+ "no_sync": 0.03884543292224407,
587
+ "no_flux": 0.04048626869916916,
588
+ "zero_source": 0.04719211347401142,
589
+ "delta_mse": 0.00997932255268097,
590
+ "delta_truth_power": 0.04778013005852699,
591
+ "delta_pred_power": 0.034492868930101395,
592
+ "rps": 0.11049726977944374,
593
+ "uniform_rps": 0.17458903044462204,
594
+ "budget": 5.960464477539063e-08,
595
+ "latent_rms": 0.5114507675170898
596
+ },
597
+ {
598
+ "seed": 990003,
599
+ "mse": 0.032159341499209404,
600
+ "persistence": 0.046610916033387184,
601
+ "swapped_controls": 0.0705743283033371,
602
+ "no_soc": 0.03262777999043465,
603
+ "no_sync": 0.0334945023059845,
604
+ "no_flux": 0.03531641699373722,
605
+ "zero_source": 0.04203234985470772,
606
+ "delta_mse": 0.012142256833612919,
607
+ "delta_truth_power": 0.05146479234099388,
608
+ "delta_pred_power": 0.03750743716955185,
609
+ "rps": 0.11526131629943848,
610
+ "uniform_rps": 0.17017009109258652,
611
+ "budget": 4.470348358154297e-08,
612
+ "latent_rms": 0.5184615850448608
613
+ },
614
+ {
615
+ "seed": 990004,
616
+ "mse": 0.04031115397810936,
617
+ "persistence": 0.05413947440683842,
618
+ "swapped_controls": 0.07577820867300034,
619
+ "no_soc": 0.0403816569596529,
620
+ "no_sync": 0.041738077998161316,
621
+ "no_flux": 0.04332899861037731,
622
+ "zero_source": 0.04904874041676521,
623
+ "delta_mse": 0.010794354602694511,
624
+ "delta_truth_power": 0.04554155841469765,
625
+ "delta_pred_power": 0.036186911165714264,
626
+ "rps": 0.11269081011414528,
627
+ "uniform_rps": 0.17716064304113388,
628
+ "budget": 5.960464477539063e-08,
629
+ "latent_rms": 0.516528844833374
630
+ },
631
+ {
632
+ "seed": 990005,
633
+ "mse": 0.03364427946507931,
634
+ "persistence": 0.04826424457132816,
635
+ "swapped_controls": 0.07187648862600327,
636
+ "no_soc": 0.03435592167079449,
637
+ "no_sync": 0.03474794514477253,
638
+ "no_flux": 0.03632078133523464,
639
+ "zero_source": 0.043854668736457825,
640
+ "delta_mse": 0.013461834751069546,
641
+ "delta_truth_power": 0.05389564111828804,
642
+ "delta_pred_power": 0.036030616611242294,
643
+ "rps": 0.11481084302067757,
644
+ "uniform_rps": 0.1723226010799408,
645
+ "budget": 4.470348358154297e-08,
646
+ "latent_rms": 0.5153510570526123
647
+ },
648
+ {
649
+ "seed": 990006,
650
+ "mse": 0.04355078004300594,
651
+ "persistence": 0.05899917148053646,
652
+ "swapped_controls": 0.08069268614053726,
653
+ "no_soc": 0.04466649331152439,
654
+ "no_sync": 0.04511422850191593,
655
+ "no_flux": 0.047012653201818466,
656
+ "zero_source": 0.05335756950080395,
657
+ "delta_mse": 0.01590139977633953,
658
+ "delta_truth_power": 0.05488300323486328,
659
+ "delta_pred_power": 0.035302210599184036,
660
+ "rps": 0.11730058118700981,
661
+ "uniform_rps": 0.17664389312267303,
662
+ "budget": 4.470348358154297e-08,
663
+ "latent_rms": 0.51296067237854
664
+ },
665
+ {
666
+ "seed": 990007,
667
+ "mse": 0.04293105937540531,
668
+ "persistence": 0.059475114569067955,
669
+ "swapped_controls": 0.08578832820057869,
670
+ "no_soc": 0.04417808912694454,
671
+ "no_sync": 0.04445432685315609,
672
+ "no_flux": 0.04641525074839592,
673
+ "zero_source": 0.054225245490670204,
674
+ "delta_mse": 0.016364002600312233,
675
+ "delta_truth_power": 0.061362240463495255,
676
+ "delta_pred_power": 0.04071629419922829,
677
+ "rps": 0.11797115206718445,
678
+ "uniform_rps": 0.17646077275276184,
679
+ "budget": 2.9802322387695312e-08,
680
+ "latent_rms": 0.5125128030776978
681
+ }
682
+ ]
683
+ },
684
+ "ood24": {
685
+ "mse": 0.09993453075488408,
686
+ "persistence": 0.12364626303315163,
687
+ "swapped_controls": 0.17446262016892433,
688
+ "no_soc": 0.09607633513708909,
689
+ "no_sync": 0.11564872972667217,
690
+ "no_flux": 0.13019570956627527,
691
+ "zero_source": 0.11070258791248004,
692
+ "delta_mse": 0.13411304851373038,
693
+ "delta_truth_power": 0.18754256268342337,
694
+ "delta_pred_power": 0.09562665969133377,
695
+ "rps": 0.15503312274813652,
696
+ "uniform_rps": 0.18398980547984442,
697
+ "budget": 4.718701044718424e-08,
698
+ "latent_rms": 0.505509614944458,
699
+ "n_seeds": 6,
700
+ "relative_mse_to_persistence": 0.8082292849246075,
701
+ "counterfactual_skill": 0.2848927379748104,
702
+ "swapped_control_penalty": 0.7457691435690046,
703
+ "no_soc_penalty": -0.038607232041327544,
704
+ "no_sync_penalty": 0.15724493679098095,
705
+ "no_flux_penalty": 0.3028100355583272,
706
+ "no_source_penalty": 0.10775111541782767,
707
+ "rps_vs_uniform": 0.8426180045346043,
708
+ "rows": [
709
+ {
710
+ "seed": 1090000,
711
+ "mse": 0.10746301710605621,
712
+ "persistence": 0.12473969906568527,
713
+ "swapped_controls": 0.16000216454267502,
714
+ "no_soc": 0.10342828556895256,
715
+ "no_sync": 0.12388692796230316,
716
+ "no_flux": 0.1399143487215042,
717
+ "zero_source": 0.1109621711075306,
718
+ "delta_mse": 0.15097810328006744,
719
+ "delta_truth_power": 0.1675889641046524,
720
+ "delta_pred_power": 0.08846742659807205,
721
+ "rps": 0.156878761947155,
722
+ "uniform_rps": 0.18147379904985428,
723
+ "budget": 4.470348358154297e-08,
724
+ "latent_rms": 0.5032488107681274
725
+ },
726
+ {
727
+ "seed": 1090001,
728
+ "mse": 0.10502408817410469,
729
+ "persistence": 0.1308807134628296,
730
+ "swapped_controls": 0.1809672713279724,
731
+ "no_soc": 0.10112134367227554,
732
+ "no_sync": 0.12041010335087776,
733
+ "no_flux": 0.13499776273965836,
734
+ "zero_source": 0.11701305210590363,
735
+ "delta_mse": 0.13678431510925293,
736
+ "delta_truth_power": 0.1990772783756256,
737
+ "delta_pred_power": 0.08959341794252396,
738
+ "rps": 0.15267561376094818,
739
+ "uniform_rps": 0.18366292864084244,
740
+ "budget": 4.470348358154297e-08,
741
+ "latent_rms": 0.5059146881103516
742
+ },
743
+ {
744
+ "seed": 1090002,
745
+ "mse": 0.09191963076591492,
746
+ "persistence": 0.11733333393931389,
747
+ "swapped_controls": 0.1732940822839737,
748
+ "no_soc": 0.08806978911161423,
749
+ "no_sync": 0.10749391093850136,
750
+ "no_flux": 0.12292732298374176,
751
+ "zero_source": 0.1042773462831974,
752
+ "delta_mse": 0.1249944269657135,
753
+ "delta_truth_power": 0.19023966789245605,
754
+ "delta_pred_power": 0.09750361740589142,
755
+ "rps": 0.15600146353244781,
756
+ "uniform_rps": 0.18841960281133652,
757
+ "budget": 5.960464477539063e-08,
758
+ "latent_rms": 0.5147852897644043
759
+ },
760
+ {
761
+ "seed": 1090003,
762
+ "mse": 0.10063784942030907,
763
+ "persistence": 0.12980291992425919,
764
+ "swapped_controls": 0.1963951140642166,
765
+ "no_soc": 0.09733515605330467,
766
+ "no_sync": 0.11625325307250023,
767
+ "no_flux": 0.128754410892725,
768
+ "zero_source": 0.117892786860466,
769
+ "delta_mse": 0.13062214851379395,
770
+ "delta_truth_power": 0.21331746876239777,
771
+ "delta_pred_power": 0.10881920158863068,
772
+ "rps": 0.15385771542787552,
773
+ "uniform_rps": 0.18636883050203323,
774
+ "budget": 4.470348358154297e-08,
775
+ "latent_rms": 0.5012614130973816
776
+ },
777
+ {
778
+ "seed": 1090004,
779
+ "mse": 0.10420002788305283,
780
+ "persistence": 0.12729327753186226,
781
+ "swapped_controls": 0.17257584631443024,
782
+ "no_soc": 0.10035447031259537,
783
+ "no_sync": 0.11856655776500702,
784
+ "no_flux": 0.1320655792951584,
785
+ "zero_source": 0.11509988829493523,
786
+ "delta_mse": 0.13188262283802032,
787
+ "delta_truth_power": 0.18447624146938324,
788
+ "delta_pred_power": 0.08415801078081131,
789
+ "rps": 0.15123936533927917,
790
+ "uniform_rps": 0.17933756858110428,
791
+ "budget": 4.470348358154297e-08,
792
+ "latent_rms": 0.49893349409103394
793
+ },
794
+ {
795
+ "seed": 1090005,
796
+ "mse": 0.09036257117986679,
797
+ "persistence": 0.11182763427495956,
798
+ "swapped_controls": 0.16354124248027802,
799
+ "no_soc": 0.08614896610379219,
800
+ "no_sync": 0.1072816252708435,
801
+ "no_flux": 0.12251483276486397,
802
+ "zero_source": 0.09897028282284737,
803
+ "delta_mse": 0.12941667437553406,
804
+ "delta_truth_power": 0.17055575549602509,
805
+ "delta_pred_power": 0.10521828383207321,
806
+ "rps": 0.15954581648111343,
807
+ "uniform_rps": 0.18467610329389572,
808
+ "budget": 4.470348358154297e-08,
809
+ "latent_rms": 0.5089139938354492
810
+ }
811
+ ]
812
+ },
813
+ "ood36_extrapolation": {
814
+ "mse": 0.1374976560473442,
815
+ "persistence": 0.15639387319485346,
816
+ "swapped_controls": 0.2011382319033146,
817
+ "no_soc": 0.12625923566520214,
818
+ "no_sync": 0.17245126763979593,
819
+ "no_flux": 0.19562622904777527,
820
+ "zero_source": 0.1433552335947752,
821
+ "delta_mse": 0.19040932257970175,
822
+ "delta_truth_power": 0.19442260016997656,
823
+ "delta_pred_power": 0.12326787536342938,
824
+ "rps": 0.18875505899389586,
825
+ "uniform_rps": 0.18619724611441293,
826
+ "budget": 5.21540641784668e-08,
827
+ "latent_rms": 0.4903332442045212,
828
+ "n_seeds": 6,
829
+ "relative_mse_to_persistence": 0.8791754640927258,
830
+ "counterfactual_skill": 0.02064203228825323,
831
+ "swapped_control_penalty": 0.46284844182403506,
832
+ "no_soc_penalty": -0.08173535975240453,
833
+ "no_sync_penalty": 0.2542124178496268,
834
+ "no_flux_penalty": 0.422760464952332,
835
+ "no_source_penalty": 0.042601290202460396,
836
+ "rps_vs_uniform": 1.0137371144464253,
837
+ "rows": [
838
+ {
839
+ "seed": 1190000,
840
+ "mse": 0.12927863374352455,
841
+ "persistence": 0.14796560257673264,
842
+ "swapped_controls": 0.1964510828256607,
843
+ "no_soc": 0.11937462911009789,
844
+ "no_sync": 0.1629735305905342,
845
+ "no_flux": 0.18221627175807953,
846
+ "zero_source": 0.13602575287222862,
847
+ "delta_mse": 0.17694734036922455,
848
+ "delta_truth_power": 0.18843600153923035,
849
+ "delta_pred_power": 0.12285622954368591,
850
+ "rps": 0.18345054984092712,
851
+ "uniform_rps": 0.18190918862819672,
852
+ "budget": 5.960464477539063e-08,
853
+ "latent_rms": 0.48652854561805725
854
+ },
855
+ {
856
+ "seed": 1190001,
857
+ "mse": 0.1459576040506363,
858
+ "persistence": 0.1610598787665367,
859
+ "swapped_controls": 0.19041349738836288,
860
+ "no_soc": 0.13372407853603363,
861
+ "no_sync": 0.17961855232715607,
862
+ "no_flux": 0.20588319003582,
863
+ "zero_source": 0.14792075008153915,
864
+ "delta_mse": 0.20550945401191711,
865
+ "delta_truth_power": 0.18829306960105896,
866
+ "delta_pred_power": 0.10612820088863373,
867
+ "rps": 0.19566548615694046,
868
+ "uniform_rps": 0.18861491978168488,
869
+ "budget": 4.470348358154297e-08,
870
+ "latent_rms": 0.4932929277420044
871
+ },
872
+ {
873
+ "seed": 1190002,
874
+ "mse": 0.11847260966897011,
875
+ "persistence": 0.1430327370762825,
876
+ "swapped_controls": 0.20104744285345078,
877
+ "no_soc": 0.10820155218243599,
878
+ "no_sync": 0.14994195848703384,
879
+ "no_flux": 0.17065158486366272,
880
+ "zero_source": 0.1313399001955986,
881
+ "delta_mse": 0.1412314921617508,
882
+ "delta_truth_power": 0.18527232110500336,
883
+ "delta_pred_power": 0.12110884487628937,
884
+ "rps": 0.178087055683136,
885
+ "uniform_rps": 0.18089193850755692,
886
+ "budget": 4.470348358154297e-08,
887
+ "latent_rms": 0.4901999831199646
888
+ },
889
+ {
890
+ "seed": 1190003,
891
+ "mse": 0.16063187271356583,
892
+ "persistence": 0.16636237502098083,
893
+ "swapped_controls": 0.1861133798956871,
894
+ "no_soc": 0.1492641344666481,
895
+ "no_sync": 0.2010246068239212,
896
+ "no_flux": 0.2299518957734108,
897
+ "zero_source": 0.15136384218931198,
898
+ "delta_mse": 0.2661841809749603,
899
+ "delta_truth_power": 0.195073202252388,
900
+ "delta_pred_power": 0.12207397073507309,
901
+ "rps": 0.19749044626951218,
902
+ "uniform_rps": 0.19155681133270264,
903
+ "budget": 4.470348358154297e-08,
904
+ "latent_rms": 0.49256622791290283
905
+ },
906
+ {
907
+ "seed": 1190004,
908
+ "mse": 0.1377757415175438,
909
+ "persistence": 0.15843652188777924,
910
+ "swapped_controls": 0.1999419778585434,
911
+ "no_soc": 0.12523655965924263,
912
+ "no_sync": 0.17292062938213348,
913
+ "no_flux": 0.19720672816038132,
914
+ "zero_source": 0.14452234655618668,
915
+ "delta_mse": 0.19439482688903809,
916
+ "delta_truth_power": 0.19655656814575195,
917
+ "delta_pred_power": 0.12217074632644653,
918
+ "rps": 0.1956218257546425,
919
+ "uniform_rps": 0.19092204421758652,
920
+ "budget": 5.960464477539063e-08,
921
+ "latent_rms": 0.4948771893978119
922
+ },
923
+ {
924
+ "seed": 1190005,
925
+ "mse": 0.13286947458982468,
926
+ "persistence": 0.16150612384080887,
927
+ "swapped_controls": 0.23286201059818268,
928
+ "no_soc": 0.12175446003675461,
929
+ "no_sync": 0.16822832822799683,
930
+ "no_flux": 0.18784770369529724,
931
+ "zero_source": 0.14895880967378616,
932
+ "delta_mse": 0.15818864107131958,
933
+ "delta_truth_power": 0.2129044383764267,
934
+ "delta_pred_power": 0.1452692598104477,
935
+ "rps": 0.18221499025821686,
936
+ "uniform_rps": 0.18328857421875,
937
+ "budget": 5.960464477539063e-08,
938
+ "latent_rms": 0.4845345914363861
939
+ }
940
+ ]
941
+ }
942
+ },
943
+ "notes": [
944
+ "Single trained model; all interventions are frozen-weight inference only",
945
+ "Physics-aware upwind Flux is a supplied inductive bias, not learned discovery",
946
+ "Synthetic future controls are known at issue, unlike unobserved real future weather",
947
+ "Counterfactual branches have identical history and distinct issued controls",
948
+ "36-step is beyond the 12/24-step training horizon",
949
+ "Paired-scale transfer balance only proves algebraic latent cancellation, not physical conservation",
950
+ "No ECMWF skill or global training approval from synthetic tests"
951
+ ]
952
+ }