anhtld commited on
Commit
b14e9c5
·
verified ·
1 Parent(s): a58ec3f

auto-sync 2026-07-04T06:15:18Z workspace (part 2)

Browse files
workspace/runs/ctt_selector_diagnostic_sweep/config.yaml CHANGED
@@ -6,6 +6,9 @@ metrics:
6
  - runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json
7
  - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json
8
  - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json
 
 
 
9
  - runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test/metrics.json
10
  - runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test/metrics.json
11
  - runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test/metrics.json
 
6
  - runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json
7
  - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json
8
  - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json
9
+ - runs/ctt_dominance_utility_energy_val_to_test_seed0/metrics.json
10
+ - runs/ctt_dominance_utility_energy_val_to_test_seed1/metrics.json
11
+ - runs/ctt_dominance_utility_energy_val_to_test_seed2/metrics.json
12
  - runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test/metrics.json
13
  - runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test/metrics.json
14
  - runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test/metrics.json
workspace/runs/ctt_selector_diagnostic_sweep/data_hash.txt CHANGED
@@ -1 +1 @@
1
- 2ec8fa640a2fa299e25d9a7decf951077112e2460bbb2c14deab81745b8249af
 
1
+ fe282d92b43f35c62c14a40a1cce310487bf3bc1adbba59184d729da713d2129
workspace/runs/ctt_selector_diagnostic_sweep/eval.log CHANGED
@@ -1,3 +1,3 @@
1
- num_inputs=15
2
- families=K8 per-dim trainmax,K8 tanh,K16 env_clip
3
  overall_pass=False
 
1
+ num_inputs=18
2
+ families=K8 per-dim trainmax,K8 tanh,K16 env_clip,K16 env_clip utility-energy
3
  overall_pass=False
workspace/runs/ctt_selector_diagnostic_sweep/metrics.json CHANGED
@@ -13,6 +13,7 @@
13
  "proposal_oracle_success": 0.2222222222222222,
14
  "report_type": "learned_dominance_selector_eval",
15
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_perdim_trainmax_train_to_test",
 
16
  "selected_success": 0.3333333333333333,
17
  "selector": "context/utility_margin, bonus=1.0",
18
  "selector_regret": 0.11280556699623251,
@@ -33,6 +34,7 @@
33
  "proposal_oracle_success": 0.3819444444444444,
34
  "report_type": "learned_dominance_selector_eval",
35
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test",
 
36
  "selected_success": 0.3819444444444444,
37
  "selector": "context/success, bonus=1.0",
38
  "selector_regret": 0.23532104610300544,
@@ -53,15 +55,37 @@
53
  "proposal_oracle_success": 0.5694444444444444,
54
  "report_type": "learned_dominance_selector_eval",
55
  "run_path": "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test",
 
56
  "selected_success": 0.3541666666666667,
57
  "selector": "chart_compat/utility_margin, chartcompat_obs",
58
  "selector_regret": 0.4511945542892338,
59
  "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
60
  "success_selector_gap": 0.24305555555555555,
61
  "success_support_gap": 0.2013888888888889
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62
  }
63
  ],
64
- "data_hash": "2ec8fa640a2fa299e25d9a7decf951077112e2460bbb2c14deab81745b8249af",
65
  "gates": [
66
  {
67
  "family": "K8 per-dim trainmax",
@@ -89,12 +113,24 @@
89
  "selector": "chart_compat/utility_margin, chartcompat_obs",
90
  "status": "diagnostic_only",
91
  "success_selector_gap": 0.24305555555555555
 
 
 
 
 
 
 
 
 
92
  }
93
  ],
94
  "input_metrics": [
95
  "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json",
96
  "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json",
97
  "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json",
 
 
 
98
  "runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test/metrics.json",
99
  "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test/metrics.json",
100
  "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test/metrics.json",
@@ -108,7 +144,7 @@
108
  "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_perdim_trainmax_train_to_test/metrics.json",
109
  "runs/ctt_base_context_obs_dominance_perdim_trainmax_train_to_test/metrics.json"
110
  ],
111
- "num_inputs": 15,
112
  "overall_pass": false,
113
  "report_type": "ctt_selector_diagnostic_sweep",
114
  "rows": [
@@ -125,6 +161,7 @@
125
  "proposal_oracle_success": 0.5694444444444444,
126
  "report_type": "learned_dominance_selector_eval",
127
  "run_path": "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test",
 
128
  "selected_success": 0.3541666666666667,
129
  "selector": "chart_compat/utility_margin, chartcompat_obs",
130
  "selector_regret": 0.4511945542892338,
@@ -145,6 +182,7 @@
145
  "proposal_oracle_success": 0.5694444444444444,
146
  "report_type": "dominance_calibrated_selector_eval",
147
  "run_path": "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test",
 
148
  "selected_success": 0.2777777777777778,
149
  "selector": "LCB auto",
150
  "selector_regret": 0.5323940152051768,
@@ -165,6 +203,7 @@
165
  "proposal_oracle_success": 0.5694444444444444,
166
  "report_type": "dominance_calibrated_selector_eval",
167
  "run_path": "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0",
 
168
  "selected_success": 0.2916666666666667,
169
  "selector": "LCB 0",
170
  "selector_regret": 0.5319753833795807,
@@ -172,6 +211,69 @@
172
  "success_selector_gap": 0.2986111111111111,
173
  "success_support_gap": 0.2013888888888889
174
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
175
  {
176
  "base_success": 0.3333333333333333,
177
  "calibration_ece": 0.02645474340097932,
@@ -185,6 +287,7 @@
185
  "proposal_oracle_success": 0.3819444444444444,
186
  "report_type": "learned_dominance_selector_eval",
187
  "run_path": "runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test",
 
188
  "selected_success": 0.3680555555555556,
189
  "selector": "basic/utility_margin, bonus=1.0",
190
  "selector_regret": 0.25259616218520226,
@@ -205,6 +308,7 @@
205
  "proposal_oracle_success": 0.3819444444444444,
206
  "report_type": "learned_dominance_selector_eval",
207
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test",
 
208
  "selected_success": 0.3819444444444444,
209
  "selector": "context/success, bonus=1.0",
210
  "selector_regret": 0.23532104610300544,
@@ -225,6 +329,7 @@
225
  "proposal_oracle_success": 0.3819444444444444,
226
  "report_type": "learned_dominance_selector_eval",
227
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test",
 
228
  "selected_success": 0.3402777777777778,
229
  "selector": "context_tangent/success, bonus=1.0",
230
  "selector_regret": 0.26619138512534035,
@@ -245,6 +350,7 @@
245
  "proposal_oracle_success": 0.3819444444444444,
246
  "report_type": "learned_dominance_selector_eval",
247
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_tanh_train_to_test",
 
248
  "selected_success": 0.3472222222222222,
249
  "selector": "context_tangent/utility_margin, bonus=1.0",
250
  "selector_regret": 0.2437624827079061,
@@ -265,6 +371,7 @@
265
  "proposal_oracle_success": 0.3819444444444444,
266
  "report_type": "learned_dominance_selector_eval",
267
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tanh_train_to_test",
 
268
  "selected_success": 0.375,
269
  "selector": "context/utility_margin, bonus=1.0",
270
  "selector_regret": 0.24082847581949965,
@@ -285,6 +392,7 @@
285
  "proposal_oracle_success": 0.3819444444444444,
286
  "report_type": "dominance_calibrated_selector_eval",
287
  "run_path": "runs/ctt_base_context_obs_dominance_tanh_train_to_test",
 
288
  "selected_success": 0.3194444444444444,
289
  "selector": "LCB auto",
290
  "selector_regret": 0.3232221397451617,
@@ -305,6 +413,7 @@
305
  "proposal_oracle_success": 0.2222222222222222,
306
  "report_type": "learned_dominance_selector_eval",
307
  "run_path": "runs/ctt_base_context_obs_learned_dominance_basic_perdim_trainmax_train_to_test",
 
308
  "selected_success": 0.3055555555555556,
309
  "selector": "basic/utility_margin, bonus=1.0",
310
  "selector_regret": 0.12719054139840105,
@@ -325,6 +434,7 @@
325
  "proposal_oracle_success": 0.2222222222222222,
326
  "report_type": "learned_dominance_selector_eval",
327
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_perdim_trainmax_train_to_test",
 
328
  "selected_success": 0.3333333333333333,
329
  "selector": "context/utility_margin, bonus=1.0",
330
  "selector_regret": 0.11280556699623251,
@@ -345,6 +455,7 @@
345
  "proposal_oracle_success": 0.2222222222222222,
346
  "report_type": "learned_dominance_selector_eval",
347
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_perdim_trainmax_train_to_test",
 
348
  "selected_success": 0.3263888888888889,
349
  "selector": "context/success, bonus=1.0",
350
  "selector_regret": 0.11845623132669264,
@@ -365,6 +476,7 @@
365
  "proposal_oracle_success": 0.2222222222222222,
366
  "report_type": "learned_dominance_selector_eval",
367
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_perdim_trainmax_train_to_test",
 
368
  "selected_success": 0.2986111111111111,
369
  "selector": "context_tangent/utility_margin, bonus=1.0",
370
  "selector_regret": 0.12803097097073785,
@@ -385,6 +497,7 @@
385
  "proposal_oracle_success": 0.2222222222222222,
386
  "report_type": "learned_dominance_selector_eval",
387
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_perdim_trainmax_train_to_test",
 
388
  "selected_success": 0.2777777777777778,
389
  "selector": "context_tangent/success, bonus=1.0",
390
  "selector_regret": 0.16677976430704197,
@@ -405,6 +518,7 @@
405
  "proposal_oracle_success": 0.2222222222222222,
406
  "report_type": "dominance_calibrated_selector_eval",
407
  "run_path": "runs/ctt_base_context_obs_dominance_perdim_trainmax_train_to_test",
 
408
  "selected_success": 0.3263888888888889,
409
  "selector": "LCB auto",
410
  "selector_regret": 0.11593854168636931,
@@ -415,7 +529,7 @@
415
  ],
416
  "schema_version": 1,
417
  "selection_rule": "best selected_success per diagnostic family; all candidate rows retained",
418
- "split_hash": "3281cb880470bb0158a5055eeb059e90f2af4a005944132e01504c225160d8c5",
419
  "thresholds": {
420
  "proposal_oracle_min": 0.5,
421
  "selected_min": 0.4745,
 
13
  "proposal_oracle_success": 0.2222222222222222,
14
  "report_type": "learned_dominance_selector_eval",
15
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_perdim_trainmax_train_to_test",
16
+ "seed": "pooled",
17
  "selected_success": 0.3333333333333333,
18
  "selector": "context/utility_margin, bonus=1.0",
19
  "selector_regret": 0.11280556699623251,
 
34
  "proposal_oracle_success": 0.3819444444444444,
35
  "report_type": "learned_dominance_selector_eval",
36
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test",
37
+ "seed": "pooled",
38
  "selected_success": 0.3819444444444444,
39
  "selector": "context/success, bonus=1.0",
40
  "selector_regret": 0.23532104610300544,
 
55
  "proposal_oracle_success": 0.5694444444444444,
56
  "report_type": "learned_dominance_selector_eval",
57
  "run_path": "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test",
58
+ "seed": "pooled",
59
  "selected_success": 0.3541666666666667,
60
  "selector": "chart_compat/utility_margin, chartcompat_obs",
61
  "selector_regret": 0.4511945542892338,
62
  "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
63
  "success_selector_gap": 0.24305555555555555,
64
  "success_support_gap": 0.2013888888888889
65
+ },
66
+ {
67
+ "base_success": 0.2916666666666667,
68
+ "calibration_ece": 0.025346812650147644,
69
+ "coverage": 0.1736111111111111,
70
+ "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
71
+ "fallback_rate": 0.8263888888888888,
72
+ "family": "K16 env_clip utility-energy",
73
+ "hidden_chart_oracle_success": 0.7291666666666666,
74
+ "k": 16,
75
+ "outcome_ptr": 0.5486111111111112,
76
+ "proposal_oracle_success": 0.5694444444444444,
77
+ "report_type": "dominance_calibrated_selector_eval",
78
+ "run_path": "runs/ctt_dominance_utility_energy_val_to_test_seed2",
79
+ "seed": "2",
80
+ "selected_success": 0.2777777777777778,
81
+ "selector": "checkpoint utility energy/LCB auto, seed=2",
82
+ "selector_regret": 0.5454409955467175,
83
+ "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
84
+ "success_selector_gap": 0.3055555555555556,
85
+ "success_support_gap": 0.2013888888888889
86
  }
87
  ],
88
+ "data_hash": "fe282d92b43f35c62c14a40a1cce310487bf3bc1adbba59184d729da713d2129",
89
  "gates": [
90
  {
91
  "family": "K8 per-dim trainmax",
 
113
  "selector": "chart_compat/utility_margin, chartcompat_obs",
114
  "status": "diagnostic_only",
115
  "success_selector_gap": 0.24305555555555555
116
+ },
117
+ {
118
+ "family": "K16 env_clip utility-energy",
119
+ "pass": false,
120
+ "proposal_oracle_success": 0.5694444444444444,
121
+ "selected_success": 0.2777777777777778,
122
+ "selector": "checkpoint utility energy/LCB auto, seed=2",
123
+ "status": "diagnostic_only",
124
+ "success_selector_gap": 0.3055555555555556
125
  }
126
  ],
127
  "input_metrics": [
128
  "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json",
129
  "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json",
130
  "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json",
131
+ "runs/ctt_dominance_utility_energy_val_to_test_seed0/metrics.json",
132
+ "runs/ctt_dominance_utility_energy_val_to_test_seed1/metrics.json",
133
+ "runs/ctt_dominance_utility_energy_val_to_test_seed2/metrics.json",
134
  "runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test/metrics.json",
135
  "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test/metrics.json",
136
  "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test/metrics.json",
 
144
  "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_perdim_trainmax_train_to_test/metrics.json",
145
  "runs/ctt_base_context_obs_dominance_perdim_trainmax_train_to_test/metrics.json"
146
  ],
147
+ "num_inputs": 18,
148
  "overall_pass": false,
149
  "report_type": "ctt_selector_diagnostic_sweep",
150
  "rows": [
 
161
  "proposal_oracle_success": 0.5694444444444444,
162
  "report_type": "learned_dominance_selector_eval",
163
  "run_path": "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test",
164
+ "seed": "pooled",
165
  "selected_success": 0.3541666666666667,
166
  "selector": "chart_compat/utility_margin, chartcompat_obs",
167
  "selector_regret": 0.4511945542892338,
 
182
  "proposal_oracle_success": 0.5694444444444444,
183
  "report_type": "dominance_calibrated_selector_eval",
184
  "run_path": "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test",
185
+ "seed": "pooled",
186
  "selected_success": 0.2777777777777778,
187
  "selector": "LCB auto",
188
  "selector_regret": 0.5323940152051768,
 
203
  "proposal_oracle_success": 0.5694444444444444,
204
  "report_type": "dominance_calibrated_selector_eval",
205
  "run_path": "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0",
206
+ "seed": "pooled",
207
  "selected_success": 0.2916666666666667,
208
  "selector": "LCB 0",
209
  "selector_regret": 0.5319753833795807,
 
211
  "success_selector_gap": 0.2986111111111111,
212
  "success_support_gap": 0.2013888888888889
213
  },
214
+ {
215
+ "base_success": 0.2916666666666667,
216
+ "calibration_ece": 0.015002442871876898,
217
+ "coverage": 0.125,
218
+ "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
219
+ "fallback_rate": 0.875,
220
+ "family": "K16 env_clip utility-energy",
221
+ "hidden_chart_oracle_success": 0.7291666666666666,
222
+ "k": 16,
223
+ "outcome_ptr": 0.5486111111111112,
224
+ "proposal_oracle_success": 0.5694444444444444,
225
+ "report_type": "dominance_calibrated_selector_eval",
226
+ "run_path": "runs/ctt_dominance_utility_energy_val_to_test_seed0",
227
+ "seed": "0",
228
+ "selected_success": 0.2708333333333333,
229
+ "selector": "checkpoint utility energy/LCB auto, seed=0",
230
+ "selector_regret": 0.573938843271915,
231
+ "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
232
+ "success_selector_gap": 0.3194444444444444,
233
+ "success_support_gap": 0.2013888888888889
234
+ },
235
+ {
236
+ "base_success": 0.2916666666666667,
237
+ "calibration_ece": 0.009001064879080955,
238
+ "coverage": 0.09722222222222222,
239
+ "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
240
+ "fallback_rate": 0.9027777777777778,
241
+ "family": "K16 env_clip utility-energy",
242
+ "hidden_chart_oracle_success": 0.7291666666666666,
243
+ "k": 16,
244
+ "outcome_ptr": 0.5486111111111112,
245
+ "proposal_oracle_success": 0.5694444444444444,
246
+ "report_type": "dominance_calibrated_selector_eval",
247
+ "run_path": "runs/ctt_dominance_utility_energy_val_to_test_seed1",
248
+ "seed": "1",
249
+ "selected_success": 0.2708333333333333,
250
+ "selector": "checkpoint utility energy/LCB auto, seed=1",
251
+ "selector_regret": 0.5715718468879155,
252
+ "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
253
+ "success_selector_gap": 0.3194444444444444,
254
+ "success_support_gap": 0.2013888888888889
255
+ },
256
+ {
257
+ "base_success": 0.2916666666666667,
258
+ "calibration_ece": 0.025346812650147644,
259
+ "coverage": 0.1736111111111111,
260
+ "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
261
+ "fallback_rate": 0.8263888888888888,
262
+ "family": "K16 env_clip utility-energy",
263
+ "hidden_chart_oracle_success": 0.7291666666666666,
264
+ "k": 16,
265
+ "outcome_ptr": 0.5486111111111112,
266
+ "proposal_oracle_success": 0.5694444444444444,
267
+ "report_type": "dominance_calibrated_selector_eval",
268
+ "run_path": "runs/ctt_dominance_utility_energy_val_to_test_seed2",
269
+ "seed": "2",
270
+ "selected_success": 0.2777777777777778,
271
+ "selector": "checkpoint utility energy/LCB auto, seed=2",
272
+ "selector_regret": 0.5454409955467175,
273
+ "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
274
+ "success_selector_gap": 0.3055555555555556,
275
+ "success_support_gap": 0.2013888888888889
276
+ },
277
  {
278
  "base_success": 0.3333333333333333,
279
  "calibration_ece": 0.02645474340097932,
 
287
  "proposal_oracle_success": 0.3819444444444444,
288
  "report_type": "learned_dominance_selector_eval",
289
  "run_path": "runs/ctt_base_context_obs_learned_dominance_basic_tanh_train_to_test",
290
+ "seed": "pooled",
291
  "selected_success": 0.3680555555555556,
292
  "selector": "basic/utility_margin, bonus=1.0",
293
  "selector_regret": 0.25259616218520226,
 
308
  "proposal_oracle_success": 0.3819444444444444,
309
  "report_type": "learned_dominance_selector_eval",
310
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_tanh_train_to_test",
311
+ "seed": "pooled",
312
  "selected_success": 0.3819444444444444,
313
  "selector": "context/success, bonus=1.0",
314
  "selector_regret": 0.23532104610300544,
 
329
  "proposal_oracle_success": 0.3819444444444444,
330
  "report_type": "learned_dominance_selector_eval",
331
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_tanh_train_to_test",
332
+ "seed": "pooled",
333
  "selected_success": 0.3402777777777778,
334
  "selector": "context_tangent/success, bonus=1.0",
335
  "selector_regret": 0.26619138512534035,
 
350
  "proposal_oracle_success": 0.3819444444444444,
351
  "report_type": "learned_dominance_selector_eval",
352
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_tanh_train_to_test",
353
+ "seed": "pooled",
354
  "selected_success": 0.3472222222222222,
355
  "selector": "context_tangent/utility_margin, bonus=1.0",
356
  "selector_regret": 0.2437624827079061,
 
371
  "proposal_oracle_success": 0.3819444444444444,
372
  "report_type": "learned_dominance_selector_eval",
373
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tanh_train_to_test",
374
+ "seed": "pooled",
375
  "selected_success": 0.375,
376
  "selector": "context/utility_margin, bonus=1.0",
377
  "selector_regret": 0.24082847581949965,
 
392
  "proposal_oracle_success": 0.3819444444444444,
393
  "report_type": "dominance_calibrated_selector_eval",
394
  "run_path": "runs/ctt_base_context_obs_dominance_tanh_train_to_test",
395
+ "seed": "pooled",
396
  "selected_success": 0.3194444444444444,
397
  "selector": "LCB auto",
398
  "selector_regret": 0.3232221397451617,
 
413
  "proposal_oracle_success": 0.2222222222222222,
414
  "report_type": "learned_dominance_selector_eval",
415
  "run_path": "runs/ctt_base_context_obs_learned_dominance_basic_perdim_trainmax_train_to_test",
416
+ "seed": "pooled",
417
  "selected_success": 0.3055555555555556,
418
  "selector": "basic/utility_margin, bonus=1.0",
419
  "selector_regret": 0.12719054139840105,
 
434
  "proposal_oracle_success": 0.2222222222222222,
435
  "report_type": "learned_dominance_selector_eval",
436
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_perdim_trainmax_train_to_test",
437
+ "seed": "pooled",
438
  "selected_success": 0.3333333333333333,
439
  "selector": "context/utility_margin, bonus=1.0",
440
  "selector_regret": 0.11280556699623251,
 
455
  "proposal_oracle_success": 0.2222222222222222,
456
  "report_type": "learned_dominance_selector_eval",
457
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_success_perdim_trainmax_train_to_test",
458
+ "seed": "pooled",
459
  "selected_success": 0.3263888888888889,
460
  "selector": "context/success, bonus=1.0",
461
  "selector_regret": 0.11845623132669264,
 
476
  "proposal_oracle_success": 0.2222222222222222,
477
  "report_type": "learned_dominance_selector_eval",
478
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_perdim_trainmax_train_to_test",
479
+ "seed": "pooled",
480
  "selected_success": 0.2986111111111111,
481
  "selector": "context_tangent/utility_margin, bonus=1.0",
482
  "selector_regret": 0.12803097097073785,
 
497
  "proposal_oracle_success": 0.2222222222222222,
498
  "report_type": "learned_dominance_selector_eval",
499
  "run_path": "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_perdim_trainmax_train_to_test",
500
+ "seed": "pooled",
501
  "selected_success": 0.2777777777777778,
502
  "selector": "context_tangent/success, bonus=1.0",
503
  "selector_regret": 0.16677976430704197,
 
518
  "proposal_oracle_success": 0.2222222222222222,
519
  "report_type": "dominance_calibrated_selector_eval",
520
  "run_path": "runs/ctt_base_context_obs_dominance_perdim_trainmax_train_to_test",
521
+ "seed": "pooled",
522
  "selected_success": 0.3263888888888889,
523
  "selector": "LCB auto",
524
  "selector_regret": 0.11593854168636931,
 
529
  ],
530
  "schema_version": 1,
531
  "selection_rule": "best selected_success per diagnostic family; all candidate rows retained",
532
+ "split_hash": "08acb82a67e5b95e3e8339cf57cfcb2695aca7e8f0b794889e754059bdd0b6b6",
533
  "thresholds": {
534
  "proposal_oracle_min": 0.5,
535
  "selected_min": 0.4745,
workspace/runs/ctt_selector_diagnostic_sweep/metrics_by_seed.json CHANGED
@@ -1,22 +1,42 @@
1
  {
2
- "16": {
3
  "base_success": 0.2916666666666667,
4
- "calibration_ece": 0.11388905660618416,
5
- "coverage": 0.3009259259259259,
6
  "outcome_ptr": 0.5486111111111112,
7
  "proposal_oracle_success": 0.5694444444444444,
8
- "selected_success": 0.3078703703703704,
9
- "success_selector_gap": 0.28009259259259256,
10
- "success_support_gap": 0.20138888888888892
11
  },
12
- "8": {
13
- "base_success": 0.3333333333333333,
14
- "calibration_ece": 0.04109681127667372,
15
- "coverage": 0.21122685185185186,
16
- "outcome_ptr": 0.38888888888888884,
17
- "proposal_oracle_success": 0.30208333333333337,
18
- "selected_success": 0.3333333333333333,
19
- "success_selector_gap": 0.10474537037037035,
20
- "success_support_gap": 0.4409722222222223
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  }
22
  }
 
1
  {
2
+ "0": {
3
  "base_success": 0.2916666666666667,
4
+ "calibration_ece": 0.015002442871876898,
5
+ "coverage": 0.125,
6
  "outcome_ptr": 0.5486111111111112,
7
  "proposal_oracle_success": 0.5694444444444444,
8
+ "selected_success": 0.2708333333333333,
9
+ "success_selector_gap": 0.3194444444444444,
10
+ "success_support_gap": 0.2013888888888889
11
  },
12
+ "1": {
13
+ "base_success": 0.2916666666666667,
14
+ "calibration_ece": 0.009001064879080955,
15
+ "coverage": 0.09722222222222222,
16
+ "outcome_ptr": 0.5486111111111112,
17
+ "proposal_oracle_success": 0.5694444444444444,
18
+ "selected_success": 0.2708333333333333,
19
+ "success_selector_gap": 0.3194444444444444,
20
+ "success_support_gap": 0.2013888888888889
21
+ },
22
+ "2": {
23
+ "base_success": 0.2916666666666667,
24
+ "calibration_ece": 0.025346812650147644,
25
+ "coverage": 0.1736111111111111,
26
+ "outcome_ptr": 0.5486111111111112,
27
+ "proposal_oracle_success": 0.5694444444444444,
28
+ "selected_success": 0.2777777777777778,
29
+ "success_selector_gap": 0.3055555555555556,
30
+ "success_support_gap": 0.2013888888888889
31
+ },
32
+ "pooled": {
33
+ "base_success": 0.325,
34
+ "calibration_ece": 0.055655260342575806,
35
+ "coverage": 0.22916666666666669,
36
+ "outcome_ptr": 0.42083333333333317,
37
+ "proposal_oracle_success": 0.3555555555555557,
38
+ "selected_success": 0.3282407407407407,
39
+ "success_selector_gap": 0.13981481481481478,
40
+ "success_support_gap": 0.3930555555555556
41
  }
42
  }
workspace/runs/ctt_selector_diagnostic_sweep/metrics_by_task.json CHANGED
@@ -9,6 +9,16 @@
9
  "success_selector_gap": 0.28009259259259256,
10
  "success_support_gap": 0.20138888888888892
11
  },
 
 
 
 
 
 
 
 
 
 
12
  "K8 per-dim trainmax": {
13
  "base_success": 0.3333333333333333,
14
  "calibration_ece": 0.03722925805679559,
 
9
  "success_selector_gap": 0.28009259259259256,
10
  "success_support_gap": 0.20138888888888892
11
  },
12
+ "K16 env_clip utility-energy": {
13
+ "base_success": 0.2916666666666667,
14
+ "calibration_ece": 0.0164501068003685,
15
+ "coverage": 0.13194444444444445,
16
+ "outcome_ptr": 0.5486111111111112,
17
+ "proposal_oracle_success": 0.5694444444444444,
18
+ "selected_success": 0.27314814814814814,
19
+ "success_selector_gap": 0.3148148148148148,
20
+ "success_support_gap": 0.20138888888888892
21
+ },
22
  "K8 per-dim trainmax": {
23
  "base_success": 0.3333333333333333,
24
  "calibration_ece": 0.03722925805679559,
workspace/runs/ctt_selector_diagnostic_sweep/split_hash.txt CHANGED
@@ -1 +1 @@
1
- 3281cb880470bb0158a5055eeb059e90f2af4a005944132e01504c225160d8c5
 
1
+ 08acb82a67e5b95e3e8339cf57cfcb2695aca7e8f0b794889e754059bdd0b6b6
workspace/runs/ctt_selector_diagnostic_sweep/table.tex CHANGED
@@ -6,5 +6,6 @@ Family & Best selector & Base & Selected & Proposal & Coverage & Sel. gap & Supp
6
  K8 per-dim trainmax & context/utility\_margin, bonus=1.0 & 0.3333 & 0.3333 & 0.2222 & 0.4236 & 0.0625 & 0.5139 \\
7
  K8 tanh & context/success, bonus=1.0 & 0.3333 & 0.3819 & 0.3819 & 0.1806 & 0.1250 & 0.3681 \\
8
  K16 env\_clip & chart\_compat/utility\_margin, chartcompat\_obs & 0.2917 & 0.3542 & 0.5694 & 0.6597 & 0.2431 & 0.2014 \\
 
9
  \bottomrule
10
  \end{tabular}
 
6
  K8 per-dim trainmax & context/utility\_margin, bonus=1.0 & 0.3333 & 0.3333 & 0.2222 & 0.4236 & 0.0625 & 0.5139 \\
7
  K8 tanh & context/success, bonus=1.0 & 0.3333 & 0.3819 & 0.3819 & 0.1806 & 0.1250 & 0.3681 \\
8
  K16 env\_clip & chart\_compat/utility\_margin, chartcompat\_obs & 0.2917 & 0.3542 & 0.5694 & 0.6597 & 0.2431 & 0.2014 \\
9
+ K16 env\_clip utility-energy & checkpoint utility energy/LCB auto, seed=2 & 0.2917 & 0.2778 & 0.5694 & 0.1736 & 0.3056 & 0.2014 \\
10
  \bottomrule
11
  \end{tabular}
workspace/runs/paper_ctt_audit/audit.json CHANGED
@@ -243,35 +243,35 @@
243
  "status": "pass"
244
  },
245
  {
246
- "line": 1067,
247
  "raw": "../runs/ctt_selector_diagnostic_sweep/table",
248
  "resolved": "runs/ctt_selector_diagnostic_sweep/table.tex",
249
  "run_dir": "runs/ctt_selector_diagnostic_sweep",
250
  "status": "pass"
251
  },
252
  {
253
- "line": 1089,
254
  "raw": "../runs/ctt_outcome_acceptance_gate/table",
255
  "resolved": "runs/ctt_outcome_acceptance_gate/table.tex",
256
  "run_dir": "runs/ctt_outcome_acceptance_gate",
257
  "status": "pass"
258
  },
259
  {
260
- "line": 1120,
261
  "raw": "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table",
262
  "resolved": "runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex",
263
  "run_dir": "runs/ctt_base_context_obs_learned_dominance_train_to_test",
264
  "status": "pass"
265
  },
266
  {
267
- "line": 1144,
268
  "raw": "../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
269
  "resolved": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
270
  "run_dir": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test",
271
  "status": "pass"
272
  },
273
  {
274
- "line": 1251,
275
  "raw": "../runs/paper_ctt_audit/table",
276
  "resolved": "runs/paper_ctt_audit/table.tex",
277
  "run_dir": "runs/paper_ctt_audit",
 
243
  "status": "pass"
244
  },
245
  {
246
+ "line": 1068,
247
  "raw": "../runs/ctt_selector_diagnostic_sweep/table",
248
  "resolved": "runs/ctt_selector_diagnostic_sweep/table.tex",
249
  "run_dir": "runs/ctt_selector_diagnostic_sweep",
250
  "status": "pass"
251
  },
252
  {
253
+ "line": 1092,
254
  "raw": "../runs/ctt_outcome_acceptance_gate/table",
255
  "resolved": "runs/ctt_outcome_acceptance_gate/table.tex",
256
  "run_dir": "runs/ctt_outcome_acceptance_gate",
257
  "status": "pass"
258
  },
259
  {
260
+ "line": 1123,
261
  "raw": "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table",
262
  "resolved": "runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex",
263
  "run_dir": "runs/ctt_base_context_obs_learned_dominance_train_to_test",
264
  "status": "pass"
265
  },
266
  {
267
+ "line": 1147,
268
  "raw": "../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
269
  "resolved": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
270
  "run_dir": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test",
271
  "status": "pass"
272
  },
273
  {
274
+ "line": 1254,
275
  "raw": "../runs/paper_ctt_audit/table",
276
  "resolved": "runs/paper_ctt_audit/table.tex",
277
  "run_dir": "runs/paper_ctt_audit",
workspace/runs/paper_ctt_audit/metrics.json CHANGED
@@ -243,35 +243,35 @@
243
  "status": "pass"
244
  },
245
  {
246
- "line": 1067,
247
  "raw": "../runs/ctt_selector_diagnostic_sweep/table",
248
  "resolved": "runs/ctt_selector_diagnostic_sweep/table.tex",
249
  "run_dir": "runs/ctt_selector_diagnostic_sweep",
250
  "status": "pass"
251
  },
252
  {
253
- "line": 1089,
254
  "raw": "../runs/ctt_outcome_acceptance_gate/table",
255
  "resolved": "runs/ctt_outcome_acceptance_gate/table.tex",
256
  "run_dir": "runs/ctt_outcome_acceptance_gate",
257
  "status": "pass"
258
  },
259
  {
260
- "line": 1120,
261
  "raw": "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table",
262
  "resolved": "runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex",
263
  "run_dir": "runs/ctt_base_context_obs_learned_dominance_train_to_test",
264
  "status": "pass"
265
  },
266
  {
267
- "line": 1144,
268
  "raw": "../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
269
  "resolved": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
270
  "run_dir": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test",
271
  "status": "pass"
272
  },
273
  {
274
- "line": 1251,
275
  "raw": "../runs/paper_ctt_audit/table",
276
  "resolved": "runs/paper_ctt_audit/table.tex",
277
  "run_dir": "runs/paper_ctt_audit",
 
243
  "status": "pass"
244
  },
245
  {
246
+ "line": 1068,
247
  "raw": "../runs/ctt_selector_diagnostic_sweep/table",
248
  "resolved": "runs/ctt_selector_diagnostic_sweep/table.tex",
249
  "run_dir": "runs/ctt_selector_diagnostic_sweep",
250
  "status": "pass"
251
  },
252
  {
253
+ "line": 1092,
254
  "raw": "../runs/ctt_outcome_acceptance_gate/table",
255
  "resolved": "runs/ctt_outcome_acceptance_gate/table.tex",
256
  "run_dir": "runs/ctt_outcome_acceptance_gate",
257
  "status": "pass"
258
  },
259
  {
260
+ "line": 1123,
261
  "raw": "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table",
262
  "resolved": "runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex",
263
  "run_dir": "runs/ctt_base_context_obs_learned_dominance_train_to_test",
264
  "status": "pass"
265
  },
266
  {
267
+ "line": 1147,
268
  "raw": "../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
269
  "resolved": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex",
270
  "run_dir": "runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test",
271
  "status": "pass"
272
  },
273
  {
274
+ "line": 1254,
275
  "raw": "../runs/paper_ctt_audit/table",
276
  "resolved": "runs/paper_ctt_audit/table.tex",
277
  "run_dir": "runs/paper_ctt_audit",