auto-sync 2026-07-04T03:59:09Z workspace (part 3)

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/command.txt

@@ -0,0 +1 @@
+python scripts/eval_learned_dominance_selector.py --calibration-input runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json --calibration-target-index data/cil_charts_rgb_refs/train/index.json --eval-input runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json --eval-target-index data/cil_charts_rgb_refs/test/index.json --source-index data/cil_charts_rgb_refs/train/index.json --checkpoint-template runs/ctt_residual_base_context_obs_seed{seed}/model.pt --out-dir runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test --feature-set score_chart_compat --selector-chart-feature-mode base_context_obs --target success_weighted_margin --success-bonus 2.0 --threshold-scope task --k 16 --bootstrap-samples 300 --no-markdown-report

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/config.yaml

@@ -0,0 +1,18 @@
+bootstrap_samples: 300
+calibration_input: runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json
+calibration_target_index: data/cil_charts_rgb_refs/train/index.json
+checkpoint_template: runs/ctt_residual_base_context_obs_seed{seed}/model.pt
+eval_input: runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+eval_target_index: data/cil_charts_rgb_refs/test/index.json
+feature_set: score_chart_compat
+fit_objective: pointwise
+k: 16
+no_markdown_report: True
+out_dir: runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test
+pairwise_weight: 1.0
+ridge_lambdas: 0,0.01,0.1,1,10,100
+selector_chart_feature_mode: base_context_obs
+source_index: data/cil_charts_rgb_refs/train/index.json
+success_bonus: 2.0
+target: success_weighted_margin
+threshold_scope: task

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/data_hash.txt

@@ -0,0 +1,7 @@
+{
+  "calibration_input": "d79e275b896eb98caa128470d4614b76960cab12407faa157c583224be7abe80",
+  "calibration_target_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85",
+  "eval_input": "541d5a928da5f7ee04ca3dea570e04208830dffde431edfe5b5e38b692634045",
+  "eval_target_index": "dee1b5225485df6624747dd4c5074bcf514685b54017d05e698d2051638fbf36",
+  "source_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85"
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/eval.log

@@ -0,0 +1,3 @@
+evaluated learned fallback rule on held-out measured rollout rows
+eval_input=runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+num_eval_rows=144

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/git_hash.txt

@@ -0,0 +1 @@
+14e81292e2004925f81e0bb4810f6f144bfd87b7

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/metrics_by_task.json

@@ -0,0 +1,137 @@
+{
+  "LiftPegUpright-v1": {
+    "base_success": 0.2,
+    "base_utility": 0.9498705387115478,
+    "coverage": 0.4666666666666667,
+    "execute_generated": 0.4666666666666667,
+    "fallback_rate": 0.5333333333333333,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.5333333333333333,
+    "pairwise_causal_calibration_accuracy": 0.5492625405285252,
+    "pairwise_causal_calibration_confidence": 0.5651094024103738,
+    "pairwise_causal_calibration_ece": 0.11881236057424142,
+    "pairwise_causal_calibration_pairs": 94.8,
+    "predicted_margin": 0.8862852359573267,
+    "proposal_oracle_success": 0.6,
+    "proposal_oracle_utility": 1.4122399787108104,
+    "selected_candidate_index": 5.333333333333333,
+    "selected_success": 0.3333333333333333,
+    "selected_success_gain_over_base": 0.13333333333333333,
+    "selected_utility": 1.088020184636116,
+    "selected_utility_gain_over_base": 0.1381496459245682,
+    "selector_regret": 0.3509542107582092,
+    "success_selector_gap": 0.26666666666666666,
+    "success_support_gap": 0.4,
+    "support_gap": 0.5877600212891897,
+    "tau": 0.87335466614541
+  },
+  "PickCube-v1": {
+    "base_success": 0.0,
+    "base_utility": 0.5839591026306152,
+    "coverage": 0.0,
+    "execute_generated": 0.0,
+    "fallback_rate": 1.0,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.0,
+    "pairwise_causal_calibration_accuracy": 0.5277777777777778,
+    "pairwise_causal_calibration_confidence": 0.5553133321393634,
+    "pairwise_causal_calibration_ece": 0.10657979977343184,
+    "pairwise_causal_calibration_pairs": 36.0,
+    "predicted_margin": 0.5842980964012888,
+    "proposal_oracle_success": 0.0,
+    "proposal_oracle_utility": 0.13425676027933756,
+    "selected_candidate_index": 7.333333333333333,
+    "selected_success": 0.0,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.5839591026306152,
+    "selected_utility_gain_over_base": 0.0,
+    "selector_regret": 0.0,
+    "success_selector_gap": 0.0,
+    "success_support_gap": 1.0,
+    "support_gap": 1.8657432397206624,
+    "tau": 1.2894892762347252
+  },
+  "PullCube-v1": {
+    "base_success": 0.36,
+    "base_utility": 0.7751704612234608,
+    "coverage": 0.6533333333333333,
+    "execute_generated": 0.6533333333333333,
+    "fallback_rate": 0.3466666666666667,
+    "hidden_chart_oracle_success": 0.52,
+    "hidden_chart_oracle_utility": 1.1467878860235214,
+    "outcome_ptr": 0.4666666666666667,
+    "pairwise_causal_calibration_accuracy": 0.5216143238307843,
+    "pairwise_causal_calibration_confidence": 0.5584521161129652,
+    "pairwise_causal_calibration_ece": 0.13244853449538532,
+    "pairwise_causal_calibration_pairs": 92.32,
+    "predicted_margin": 0.6759761060180944,
+    "proposal_oracle_success": 0.5333333333333333,
+    "proposal_oracle_utility": 1.1093241602430741,
+    "selected_candidate_index": 5.986666666666666,
+    "selected_success": 0.29333333333333333,
+    "selected_success_gain_over_base": -0.06666666666666667,
+    "selected_utility": 0.6341658942727372,
+    "selected_utility_gain_over_base": -0.14100456695072353,
+    "selector_regret": 0.4787018513850247,
+    "success_selector_gap": 0.24,
+    "success_support_gap": 0.02666666666666667,
+    "support_gap": 0.1075624164690574,
+    "tau": 0.5825884328057477
+  },
+  "PushCube-v1": {
+    "base_success": 0.16666666666666666,
+    "base_utility": 0.3924041787783305,
+    "coverage": 0.7777777777777778,
+    "execute_generated": 0.7777777777777778,
+    "fallback_rate": 0.2222222222222222,
+    "hidden_chart_oracle_success": 0.8333333333333334,
+    "hidden_chart_oracle_utility": 1.7015800600250561,
+    "outcome_ptr": 0.8333333333333334,
+    "pairwise_causal_calibration_accuracy": 0.5465423944585421,
+    "pairwise_causal_calibration_confidence": 0.5419772241344875,
+    "pairwise_causal_calibration_ece": 0.11799694975778399,
+    "pairwise_causal_calibration_pairs": 97.44444444444444,
+    "predicted_margin": 1.3829754161601442,
+    "proposal_oracle_success": 0.8888888888888888,
+    "proposal_oracle_utility": 1.7857814149724112,
+    "selected_candidate_index": 3.7222222222222223,
+    "selected_success": 0.5555555555555556,
+    "selected_success_gain_over_base": 0.3888888888888889,
+    "selected_utility": 1.162042656706439,
+    "selected_utility_gain_over_base": 0.7696384779281087,
+    "selector_regret": 0.6237387582659721,
+    "success_selector_gap": 0.3333333333333333,
+    "success_support_gap": 0.1111111111111111,
+    "support_gap": 0.21421858502758873,
+    "tau": 0.8468573198810989
+  },
+  "StackCube-v1": {
+    "base_success": 0.2727272727272727,
+    "base_utility": 0.7763947045261209,
+    "coverage": 0.2727272727272727,
+    "execute_generated": 0.2727272727272727,
+    "fallback_rate": 0.7272727272727273,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.6363636363636364,
+    "pairwise_causal_calibration_accuracy": 0.5043423274143793,
+    "pairwise_causal_calibration_confidence": 0.5567990222574275,
+    "pairwise_causal_calibration_ece": 0.13094581844275163,
+    "pairwise_causal_calibration_pairs": 67.45454545454545,
+    "predicted_margin": 0.7229237046506615,
+    "proposal_oracle_success": 0.5151515151515151,
+    "proposal_oracle_utility": 1.1567951707225856,
+    "selected_candidate_index": 5.393939393939394,
+    "selected_success": 0.2727272727272727,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.7495684027671814,
+    "selected_utility_gain_over_base": -0.02682630175893957,
+    "selector_regret": 0.41081975310137775,
+    "success_selector_gap": 0.24242424242424243,
+    "success_support_gap": 0.48484848484848486,
+    "support_gap": 0.8432048292774142,
+    "tau": 1.0718654827809493
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/split_hash.txt

@@ -0,0 +1,14 @@
+{
+  "calibration_target_index": {
+    "content_hash": "67f7b4a692f7c7e71da24378dc71e3399c5d35afc14c494869eba6087b765c42",
+    "retrieval_index_allowed": true,
+    "split": "train",
+    "split_hash": "867a94a1c67686c082765671a7bf8d74a9d633085e1b2619c83197c60bf58372"
+  },
+  "eval_target_index": {
+    "content_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+    "retrieval_index_allowed": false,
+    "split": "test",
+    "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f"
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/table.tex

@@ -0,0 +1,8 @@
+% Auto-generated by scripts/eval_learned_dominance_selector.py
+\begin{tabular}{lrrrrrrrrr}
+\toprule
+Rows & Coverage & Fallback & Base succ. & Selected succ. & Oracle succ. & OutcomePTR & Succ. support gap & Succ. selector gap & Cal. ECE \\
+\midrule
+144 & 0.5486 & 0.4514 & 0.2917 & 0.3194 & 0.5694 & 0.5486 & 0.2014 & 0.2500 & 0.0302 \\
+\bottomrule
+\end{tabular}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_successbonus2_task_envclip_k16_train_to_test/train.log

@@ -0,0 +1,5 @@
+trained ridge dominance calibrator on calibration measured rows only
+fit_objective=pointwise
+selected_lambda=10.0
+tau={"LiftPegUpright-v1": 0.873355, "PickCube-v1": 1.289489, "PullCube-v1": 0.582588, "PushCube-v1": 0.846857, "StackCube-v1": 1.071865, "__global__": 0.933426}
+calibration_input=runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/command.txt

@@ -0,0 +1 @@
+python scripts/eval_learned_dominance_selector.py --calibration-input runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json --calibration-target-index data/cil_charts_rgb_refs/train/index.json --eval-input runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json --eval-target-index data/cil_charts_rgb_refs/test/index.json --source-index data/cil_charts_rgb_refs/train/index.json --checkpoint-template runs/ctt_residual_base_context_obs_seed{seed}/model.pt --out-dir runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test --feature-set score_chart_compat --selector-chart-feature-mode base_context_obs --target utility_margin --threshold-scope task --k 16 --bootstrap-samples 300 --no-markdown-report

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/config.yaml

@@ -0,0 +1,18 @@
+bootstrap_samples: 300
+calibration_input: runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json
+calibration_target_index: data/cil_charts_rgb_refs/train/index.json
+checkpoint_template: runs/ctt_residual_base_context_obs_seed{seed}/model.pt
+eval_input: runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+eval_target_index: data/cil_charts_rgb_refs/test/index.json
+feature_set: score_chart_compat
+fit_objective: pointwise
+k: 16
+no_markdown_report: True
+out_dir: runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test
+pairwise_weight: 1.0
+ridge_lambdas: 0,0.01,0.1,1,10,100
+selector_chart_feature_mode: base_context_obs
+source_index: data/cil_charts_rgb_refs/train/index.json
+success_bonus: 1.0
+target: utility_margin
+threshold_scope: task

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/data_hash.txt

@@ -0,0 +1,7 @@
+{
+  "calibration_input": "d79e275b896eb98caa128470d4614b76960cab12407faa157c583224be7abe80",
+  "calibration_target_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85",
+  "eval_input": "541d5a928da5f7ee04ca3dea570e04208830dffde431edfe5b5e38b692634045",
+  "eval_target_index": "dee1b5225485df6624747dd4c5074bcf514685b54017d05e698d2051638fbf36",
+  "source_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85"
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/eval.log

@@ -0,0 +1,3 @@
+evaluated learned fallback rule on held-out measured rollout rows
+eval_input=runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+num_eval_rows=144

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/git_hash.txt

@@ -0,0 +1 @@
+14e81292e2004925f81e0bb4810f6f144bfd87b7

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/metrics_by_task.json

@@ -0,0 +1,137 @@
+{
+  "LiftPegUpright-v1": {
+    "base_success": 0.2,
+    "base_utility": 0.9498705387115478,
+    "coverage": 0.6666666666666666,
+    "execute_generated": 0.6666666666666666,
+    "fallback_rate": 0.3333333333333333,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.5333333333333333,
+    "pairwise_causal_calibration_accuracy": 0.5494360697557629,
+    "pairwise_causal_calibration_confidence": 0.5321483671612611,
+    "pairwise_causal_calibration_ece": 0.08779692862966537,
+    "pairwise_causal_calibration_pairs": 94.8,
+    "predicted_margin": 0.09934494198504117,
+    "proposal_oracle_success": 0.6,
+    "proposal_oracle_utility": 1.4122399787108104,
+    "selected_candidate_index": 7.2,
+    "selected_success": 0.26666666666666666,
+    "selected_success_gain_over_base": 0.06666666666666667,
+    "selected_utility": 0.9396337101856868,
+    "selected_utility_gain_over_base": -0.010236828525861105,
+    "selector_regret": 0.47343848447004955,
+    "success_selector_gap": 0.3333333333333333,
+    "success_support_gap": 0.4,
+    "support_gap": 0.5877600212891897,
+    "tau": 0.028235521236802175
+  },
+  "PickCube-v1": {
+    "base_success": 0.0,
+    "base_utility": 0.5839591026306152,
+    "coverage": 0.0,
+    "execute_generated": 0.0,
+    "fallback_rate": 1.0,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.0,
+    "pairwise_causal_calibration_accuracy": 0.3796296296296296,
+    "pairwise_causal_calibration_confidence": 0.5397879799473912,
+    "pairwise_causal_calibration_ece": 0.16015835031776168,
+    "pairwise_causal_calibration_pairs": 36.0,
+    "predicted_margin": -0.2678354586918643,
+    "proposal_oracle_success": 0.0,
+    "proposal_oracle_utility": 0.13425676027933756,
+    "selected_candidate_index": 1.3333333333333333,
+    "selected_success": 0.0,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.5839591026306152,
+    "selected_utility_gain_over_base": 0.0,
+    "selector_regret": 0.0,
+    "success_selector_gap": 0.0,
+    "success_support_gap": 1.0,
+    "support_gap": 1.8657432397206624,
+    "tau": 0.5025584143631503
+  },
+  "PullCube-v1": {
+    "base_success": 0.36,
+    "base_utility": 0.7751704612234608,
+    "coverage": 0.8133333333333334,
+    "execute_generated": 0.8133333333333334,
+    "fallback_rate": 0.18666666666666668,
+    "hidden_chart_oracle_success": 0.52,
+    "hidden_chart_oracle_utility": 1.1467878860235214,
+    "outcome_ptr": 0.4666666666666667,
+    "pairwise_causal_calibration_accuracy": 0.5439315887395845,
+    "pairwise_causal_calibration_confidence": 0.5315713773974821,
+    "pairwise_causal_calibration_ece": 0.1202240575596694,
+    "pairwise_causal_calibration_pairs": 92.32,
+    "predicted_margin": 0.0587216273047277,
+    "proposal_oracle_success": 0.5333333333333333,
+    "proposal_oracle_utility": 1.1093241602430741,
+    "selected_candidate_index": 6.16,
+    "selected_success": 0.32,
+    "selected_success_gain_over_base": -0.04,
+    "selected_utility": 0.6822500004371007,
+    "selected_utility_gain_over_base": -0.09292046078636,
+    "selector_regret": 0.4528969661643108,
+    "success_selector_gap": 0.22666666666666666,
+    "success_support_gap": 0.02666666666666667,
+    "support_gap": 0.1075624164690574,
+    "tau": -0.2012384152043576
+  },
+  "PushCube-v1": {
+    "base_success": 0.16666666666666666,
+    "base_utility": 0.3924041787783305,
+    "coverage": 0.7777777777777778,
+    "execute_generated": 0.7777777777777778,
+    "fallback_rate": 0.2222222222222222,
+    "hidden_chart_oracle_success": 0.8333333333333334,
+    "hidden_chart_oracle_utility": 1.7015800600250561,
+    "outcome_ptr": 0.8333333333333334,
+    "pairwise_causal_calibration_accuracy": 0.5408758312375324,
+    "pairwise_causal_calibration_confidence": 0.5154259344288218,
+    "pairwise_causal_calibration_ece": 0.09497784791093052,
+    "pairwise_causal_calibration_pairs": 97.44444444444444,
+    "predicted_margin": 0.32846837901195985,
+    "proposal_oracle_success": 0.8888888888888888,
+    "proposal_oracle_utility": 1.7857814149724112,
+    "selected_candidate_index": 5.222222222222222,
+    "selected_success": 0.5,
+    "selected_success_gain_over_base": 0.3333333333333333,
+    "selected_utility": 1.0611259018381436,
+    "selected_utility_gain_over_base": 0.6687217230598131,
+    "selector_regret": 0.7246555131342676,
+    "success_selector_gap": 0.3888888888888889,
+    "success_support_gap": 0.1111111111111111,
+    "support_gap": 0.21421858502758873,
+    "tau": 0.0742008167559442
+  },
+  "StackCube-v1": {
+    "base_success": 0.2727272727272727,
+    "base_utility": 0.7763947045261209,
+    "coverage": 0.21212121212121213,
+    "execute_generated": 0.21212121212121213,
+    "fallback_rate": 0.7878787878787878,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.6363636363636364,
+    "pairwise_causal_calibration_accuracy": 0.4871933998564077,
+    "pairwise_causal_calibration_confidence": 0.5389990977485094,
+    "pairwise_causal_calibration_ece": 0.10659066999978553,
+    "pairwise_causal_calibration_pairs": 67.45454545454545,
+    "predicted_margin": 0.03761618987646351,
+    "proposal_oracle_success": 0.5151515151515151,
+    "proposal_oracle_utility": 1.1567951707225856,
+    "selected_candidate_index": 5.606060606060606,
+    "selected_success": 0.30303030303030304,
+    "selected_success_gain_over_base": 0.030303030303030304,
+    "selected_utility": 0.769150389414845,
+    "selected_utility_gain_over_base": -0.007244315111275875,
+    "selector_regret": 0.3912377664537141,
+    "success_selector_gap": 0.21212121212121213,
+    "success_support_gap": 0.48484848484848486,
+    "support_gap": 0.8432048292774142,
+    "tau": 0.13235082030965997
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/split_hash.txt

@@ -0,0 +1,14 @@
+{
+  "calibration_target_index": {
+    "content_hash": "67f7b4a692f7c7e71da24378dc71e3399c5d35afc14c494869eba6087b765c42",
+    "retrieval_index_allowed": true,
+    "split": "train",
+    "split_hash": "867a94a1c67686c082765671a7bf8d74a9d633085e1b2619c83197c60bf58372"
+  },
+  "eval_target_index": {
+    "content_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+    "retrieval_index_allowed": false,
+    "split": "test",
+    "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f"
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/table.tex

@@ -0,0 +1,8 @@
+% Auto-generated by scripts/eval_learned_dominance_selector.py
+\begin{tabular}{lrrrrrrrrr}
+\toprule
+Rows & Coverage & Fallback & Base succ. & Selected succ. & Oracle succ. & OutcomePTR & Succ. support gap & Succ. selector gap & Cal. ECE \\
+\midrule
+144 & 0.6389 & 0.3611 & 0.2917 & 0.3264 & 0.5694 & 0.5486 & 0.2014 & 0.2500 & 0.0221 \\
+\bottomrule
+\end{tabular}

workspace/runs/ctt_base_context_obs_learned_dominance_score_chartcompat_utility_task_envclip_k16_train_to_test/train.log

@@ -0,0 +1,5 @@
+trained ridge dominance calibrator on calibration measured rows only
+fit_objective=pointwise
+selected_lambda=100.0
+tau={"LiftPegUpright-v1": 0.028236, "PickCube-v1": 0.502558, "PullCube-v1": -0.201238, "PushCube-v1": 0.074201, "StackCube-v1": 0.132351, "__global__": 0.018529}
+calibration_input=runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/command.txt

@@ -0,0 +1 @@
+python scripts/eval_learned_dominance_selector.py --calibration-input runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json --calibration-target-index data/cil_charts_rgb_refs/train/index.json --eval-input runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json --eval-target-index data/cil_charts_rgb_refs/test/index.json --source-index data/cil_charts_rgb_refs/train/index.json --checkpoint-template runs/ctt_residual_base_context_obs_seed{seed}/model.pt --out-dir runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test --feature-set score_context_chart_compat --selector-chart-feature-mode base_context_obs --target success_weighted_margin --success-bonus 2.0 --threshold-scope task --k 16 --bootstrap-samples 300 --no-markdown-report

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/config.yaml

@@ -0,0 +1,18 @@
+bootstrap_samples: 300
+calibration_input: runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json
+calibration_target_index: data/cil_charts_rgb_refs/train/index.json
+checkpoint_template: runs/ctt_residual_base_context_obs_seed{seed}/model.pt
+eval_input: runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+eval_target_index: data/cil_charts_rgb_refs/test/index.json
+feature_set: score_context_chart_compat
+fit_objective: pointwise
+k: 16
+no_markdown_report: True
+out_dir: runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test
+pairwise_weight: 1.0
+ridge_lambdas: 0,0.01,0.1,1,10,100
+selector_chart_feature_mode: base_context_obs
+source_index: data/cil_charts_rgb_refs/train/index.json
+success_bonus: 2.0
+target: success_weighted_margin
+threshold_scope: task

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/data_hash.txt

@@ -0,0 +1,7 @@
+{
+  "calibration_input": "d79e275b896eb98caa128470d4614b76960cab12407faa157c583224be7abe80",
+  "calibration_target_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85",
+  "eval_input": "541d5a928da5f7ee04ca3dea570e04208830dffde431edfe5b5e38b692634045",
+  "eval_target_index": "dee1b5225485df6624747dd4c5074bcf514685b54017d05e698d2051638fbf36",
+  "source_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85"
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/eval.log

@@ -0,0 +1,3 @@
+evaluated learned fallback rule on held-out measured rollout rows
+eval_input=runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+num_eval_rows=144

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/git_hash.txt

@@ -0,0 +1 @@
+14e81292e2004925f81e0bb4810f6f144bfd87b7

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/metrics_by_task.json

@@ -0,0 +1,137 @@
+{
+  "LiftPegUpright-v1": {
+    "base_success": 0.2,
+    "base_utility": 0.9498705387115478,
+    "coverage": 0.7333333333333333,
+    "execute_generated": 0.7333333333333333,
+    "fallback_rate": 0.26666666666666666,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.5333333333333333,
+    "pairwise_causal_calibration_accuracy": 0.5688090610724882,
+    "pairwise_causal_calibration_confidence": 0.5640968449927838,
+    "pairwise_causal_calibration_ece": 0.10494933398158825,
+    "pairwise_causal_calibration_pairs": 94.8,
+    "predicted_margin": 0.6688100344694228,
+    "proposal_oracle_success": 0.6,
+    "proposal_oracle_utility": 1.4122399787108104,
+    "selected_candidate_index": 6.0,
+    "selected_success": 0.2,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.8018400182326635,
+    "selected_utility_gain_over_base": -0.14803052047888438,
+    "selector_regret": 0.6113315612077713,
+    "success_selector_gap": 0.4,
+    "success_support_gap": 0.4,
+    "support_gap": 0.5877600212891897,
+    "tau": 0.49469300263486815
+  },
+  "PickCube-v1": {
+    "base_success": 0.0,
+    "base_utility": 0.5839591026306152,
+    "coverage": 0.0,
+    "execute_generated": 0.0,
+    "fallback_rate": 1.0,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.0,
+    "pairwise_causal_calibration_accuracy": 0.5555555555555555,
+    "pairwise_causal_calibration_confidence": 0.5748727583608265,
+    "pairwise_causal_calibration_ece": 0.12483284745128205,
+    "pairwise_causal_calibration_pairs": 36.0,
+    "predicted_margin": 0.5222025116102257,
+    "proposal_oracle_success": 0.0,
+    "proposal_oracle_utility": 0.13425676027933756,
+    "selected_candidate_index": 7.333333333333333,
+    "selected_success": 0.0,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.5839591026306152,
+    "selected_utility_gain_over_base": 0.0,
+    "selector_regret": 0.0,
+    "success_selector_gap": 0.0,
+    "success_support_gap": 1.0,
+    "support_gap": 1.8657432397206624,
+    "tau": 0.8901452015560949
+  },
+  "PullCube-v1": {
+    "base_success": 0.36,
+    "base_utility": 0.7751704612234608,
+    "coverage": 0.8133333333333334,
+    "execute_generated": 0.8133333333333334,
+    "fallback_rate": 0.18666666666666668,
+    "hidden_chart_oracle_success": 0.52,
+    "hidden_chart_oracle_utility": 1.1467878860235214,
+    "outcome_ptr": 0.4666666666666667,
+    "pairwise_causal_calibration_accuracy": 0.5253206716900769,
+    "pairwise_causal_calibration_confidence": 0.5764914402555266,
+    "pairwise_causal_calibration_ece": 0.13706044945766013,
+    "pairwise_causal_calibration_pairs": 92.32,
+    "predicted_margin": 0.8087553078530508,
+    "proposal_oracle_success": 0.5333333333333333,
+    "proposal_oracle_utility": 1.1093241602430741,
+    "selected_candidate_index": 5.946666666666666,
+    "selected_success": 0.3466666666666667,
+    "selected_success_gain_over_base": -0.013333333333333334,
+    "selected_utility": 0.7315792678203433,
+    "selected_utility_gain_over_base": -0.04359119340311736,
+    "selector_regret": 0.3792251957549403,
+    "success_selector_gap": 0.18666666666666668,
+    "success_support_gap": 0.02666666666666667,
+    "support_gap": 0.1075624164690574,
+    "tau": 0.3663750360591525
+  },
+  "PushCube-v1": {
+    "base_success": 0.16666666666666666,
+    "base_utility": 0.3924041787783305,
+    "coverage": 0.7222222222222222,
+    "execute_generated": 0.7222222222222222,
+    "fallback_rate": 0.2777777777777778,
+    "hidden_chart_oracle_success": 0.8333333333333334,
+    "hidden_chart_oracle_utility": 1.7015800600250561,
+    "outcome_ptr": 0.8333333333333334,
+    "pairwise_causal_calibration_accuracy": 0.5564552891415517,
+    "pairwise_causal_calibration_confidence": 0.5440196808663499,
+    "pairwise_causal_calibration_ece": 0.09639707450334344,
+    "pairwise_causal_calibration_pairs": 97.44444444444444,
+    "predicted_margin": 1.5178990886097534,
+    "proposal_oracle_success": 0.8888888888888888,
+    "proposal_oracle_utility": 1.7857814149724112,
+    "selected_candidate_index": 3.6666666666666665,
+    "selected_success": 0.6111111111111112,
+    "selected_success_gain_over_base": 0.4444444444444444,
+    "selected_utility": 1.2593486458063126,
+    "selected_utility_gain_over_base": 0.8669444670279821,
+    "selector_regret": 0.5264327691660987,
+    "success_selector_gap": 0.2777777777777778,
+    "success_support_gap": 0.1111111111111111,
+    "support_gap": 0.21421858502758873,
+    "tau": 1.311950902009979
+  },
+  "StackCube-v1": {
+    "base_success": 0.2727272727272727,
+    "base_utility": 0.7763947045261209,
+    "coverage": 0.09090909090909091,
+    "execute_generated": 0.09090909090909091,
+    "fallback_rate": 0.9090909090909091,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.6363636363636364,
+    "pairwise_causal_calibration_accuracy": 0.5212002507574325,
+    "pairwise_causal_calibration_confidence": 0.5650793632911131,
+    "pairwise_causal_calibration_ece": 0.12478007178656221,
+    "pairwise_causal_calibration_pairs": 67.45454545454545,
+    "predicted_margin": 0.8903319056592717,
+    "proposal_oracle_success": 0.5151515151515151,
+    "proposal_oracle_utility": 1.1567951707225856,
+    "selected_candidate_index": 6.333333333333333,
+    "selected_success": 0.2727272727272727,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.7337339899756692,
+    "selected_utility_gain_over_base": -0.0426607145504518,
+    "selector_regret": 0.42665416589289,
+    "success_selector_gap": 0.24242424242424243,
+    "success_support_gap": 0.48484848484848486,
+    "support_gap": 0.8432048292774142,
+    "tau": 1.439329801233069
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/split_hash.txt

@@ -0,0 +1,14 @@
+{
+  "calibration_target_index": {
+    "content_hash": "67f7b4a692f7c7e71da24378dc71e3399c5d35afc14c494869eba6087b765c42",
+    "retrieval_index_allowed": true,
+    "split": "train",
+    "split_hash": "867a94a1c67686c082765671a7bf8d74a9d633085e1b2619c83197c60bf58372"
+  },
+  "eval_target_index": {
+    "content_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+    "retrieval_index_allowed": false,
+    "split": "test",
+    "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f"
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/table.tex

@@ -0,0 +1,8 @@
+% Auto-generated by scripts/eval_learned_dominance_selector.py
+\begin{tabular}{lrrrrrrrrr}
+\toprule
+Rows & Coverage & Fallback & Base succ. & Selected succ. & Oracle succ. & OutcomePTR & Succ. support gap & Succ. selector gap & Cal. ECE \\
+\midrule
+144 & 0.6111 & 0.3889 & 0.2917 & 0.3403 & 0.5694 & 0.5486 & 0.2014 & 0.2292 & 0.0365 \\
+\bottomrule
+\end{tabular}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_successbonus2_task_envclip_k16_train_to_test/train.log

@@ -0,0 +1,5 @@
+trained ridge dominance calibrator on calibration measured rows only
+fit_objective=pointwise
+selected_lambda=0.01
+tau={"LiftPegUpright-v1": 0.494693, "PickCube-v1": 0.890145, "PullCube-v1": 0.366375, "PushCube-v1": 1.311951, "StackCube-v1": 1.43933, "__global__": 1.342807}
+calibration_input=runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/command.txt

@@ -0,0 +1 @@
+python scripts/eval_learned_dominance_selector.py --calibration-input runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json --calibration-target-index data/cil_charts_rgb_refs/train/index.json --eval-input runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json --eval-target-index data/cil_charts_rgb_refs/test/index.json --source-index data/cil_charts_rgb_refs/train/index.json --checkpoint-template runs/ctt_residual_base_context_obs_seed{seed}/model.pt --out-dir runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test --feature-set score_context_chart_compat --selector-chart-feature-mode base_context_obs --target utility_margin --threshold-scope task --k 16 --bootstrap-samples 300 --no-markdown-report

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/config.yaml

@@ -0,0 +1,18 @@
+bootstrap_samples: 300
+calibration_input: runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json
+calibration_target_index: data/cil_charts_rgb_refs/train/index.json
+checkpoint_template: runs/ctt_residual_base_context_obs_seed{seed}/model.pt
+eval_input: runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+eval_target_index: data/cil_charts_rgb_refs/test/index.json
+feature_set: score_context_chart_compat
+fit_objective: pointwise
+k: 16
+no_markdown_report: True
+out_dir: runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test
+pairwise_weight: 1.0
+ridge_lambdas: 0,0.01,0.1,1,10,100
+selector_chart_feature_mode: base_context_obs
+source_index: data/cil_charts_rgb_refs/train/index.json
+success_bonus: 1.0
+target: utility_margin
+threshold_scope: task

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/data_hash.txt

@@ -0,0 +1,7 @@
+{
+  "calibration_input": "d79e275b896eb98caa128470d4614b76960cab12407faa157c583224be7abe80",
+  "calibration_target_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85",
+  "eval_input": "541d5a928da5f7ee04ca3dea570e04208830dffde431edfe5b5e38b692634045",
+  "eval_target_index": "dee1b5225485df6624747dd4c5074bcf514685b54017d05e698d2051638fbf36",
+  "source_index": "67973c50072aa87a3b97533b6292b08a8b4d88a95571dc22e91b6e8fd19e0d85"
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/eval.log

@@ -0,0 +1,3 @@
+evaluated learned fallback rule on held-out measured rollout rows
+eval_input=runs/ctt_base_context_obs_test_envclip_k16_rollout_comparison/combined_measured_candidates.json
+num_eval_rows=144

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/git_hash.txt

@@ -0,0 +1 @@
+14e81292e2004925f81e0bb4810f6f144bfd87b7

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/metrics_by_task.json

@@ -0,0 +1,137 @@
+{
+  "LiftPegUpright-v1": {
+    "base_success": 0.2,
+    "base_utility": 0.9498705387115478,
+    "coverage": 0.6666666666666666,
+    "execute_generated": 0.6666666666666666,
+    "fallback_rate": 0.3333333333333333,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.5333333333333333,
+    "pairwise_causal_calibration_accuracy": 0.5353055242569309,
+    "pairwise_causal_calibration_confidence": 0.5334598489552488,
+    "pairwise_causal_calibration_ece": 0.08576162490103227,
+    "pairwise_causal_calibration_pairs": 94.8,
+    "predicted_margin": 0.023658521728357904,
+    "proposal_oracle_success": 0.6,
+    "proposal_oracle_utility": 1.4122399787108104,
+    "selected_candidate_index": 7.2,
+    "selected_success": 0.26666666666666666,
+    "selected_success_gain_over_base": 0.06666666666666667,
+    "selected_utility": 0.9396337101856868,
+    "selected_utility_gain_over_base": -0.010236828525861105,
+    "selector_regret": 0.47343848447004955,
+    "success_selector_gap": 0.3333333333333333,
+    "success_support_gap": 0.4,
+    "support_gap": 0.5877600212891897,
+    "tau": -0.08577718626706465
+  },
+  "PickCube-v1": {
+    "base_success": 0.0,
+    "base_utility": 0.5839591026306152,
+    "coverage": 0.0,
+    "execute_generated": 0.0,
+    "fallback_rate": 1.0,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.0,
+    "pairwise_causal_calibration_accuracy": 0.3703703703703704,
+    "pairwise_causal_calibration_confidence": 0.5425410371755132,
+    "pairwise_causal_calibration_ece": 0.17217066680514279,
+    "pairwise_causal_calibration_pairs": 36.0,
+    "predicted_margin": -0.42692880044302434,
+    "proposal_oracle_success": 0.0,
+    "proposal_oracle_utility": 0.13425676027933756,
+    "selected_candidate_index": 1.3333333333333333,
+    "selected_success": 0.0,
+    "selected_success_gain_over_base": 0.0,
+    "selected_utility": 0.5839591026306152,
+    "selected_utility_gain_over_base": 0.0,
+    "selector_regret": 0.0,
+    "success_selector_gap": 0.0,
+    "success_support_gap": 1.0,
+    "support_gap": 1.8657432397206624,
+    "tau": 0.42252559087721425
+  },
+  "PullCube-v1": {
+    "base_success": 0.36,
+    "base_utility": 0.7751704612234608,
+    "coverage": 0.7733333333333333,
+    "execute_generated": 0.7733333333333333,
+    "fallback_rate": 0.22666666666666666,
+    "hidden_chart_oracle_success": 0.52,
+    "hidden_chart_oracle_utility": 1.1467878860235214,
+    "outcome_ptr": 0.4666666666666667,
+    "pairwise_causal_calibration_accuracy": 0.5394044361388536,
+    "pairwise_causal_calibration_confidence": 0.532007783280179,
+    "pairwise_causal_calibration_ece": 0.1182675802568201,
+    "pairwise_causal_calibration_pairs": 92.32,
+    "predicted_margin": 0.05246869761637144,
+    "proposal_oracle_success": 0.5333333333333333,
+    "proposal_oracle_utility": 1.1093241602430741,
+    "selected_candidate_index": 6.386666666666667,
+    "selected_success": 0.32,
+    "selected_success_gain_over_base": -0.04,
+    "selected_utility": 0.6831284697043399,
+    "selected_utility_gain_over_base": -0.09204199151912083,
+    "selector_regret": 0.4520184968970716,
+    "success_selector_gap": 0.22666666666666666,
+    "success_support_gap": 0.02666666666666667,
+    "support_gap": 0.1075624164690574,
+    "tau": -0.17411627944196265
+  },
+  "PushCube-v1": {
+    "base_success": 0.16666666666666666,
+    "base_utility": 0.3924041787783305,
+    "coverage": 0.7777777777777778,
+    "execute_generated": 0.7777777777777778,
+    "fallback_rate": 0.2222222222222222,
+    "hidden_chart_oracle_success": 0.8333333333333334,
+    "hidden_chart_oracle_utility": 1.7015800600250561,
+    "outcome_ptr": 0.8333333333333334,
+    "pairwise_causal_calibration_accuracy": 0.5261888915984856,
+    "pairwise_causal_calibration_confidence": 0.5156592220791224,
+    "pairwise_causal_calibration_ece": 0.10017848048000332,
+    "pairwise_causal_calibration_pairs": 97.44444444444444,
+    "predicted_margin": 0.31253125927801373,
+    "proposal_oracle_success": 0.8888888888888888,
+    "proposal_oracle_utility": 1.7857814149724112,
+    "selected_candidate_index": 4.5,
+    "selected_success": 0.5,
+    "selected_success_gain_over_base": 0.3333333333333333,
+    "selected_utility": 1.0588933833771281,
+    "selected_utility_gain_over_base": 0.6664892045987977,
+    "selector_regret": 0.726888031595283,
+    "success_selector_gap": 0.3888888888888889,
+    "success_support_gap": 0.1111111111111111,
+    "support_gap": 0.21421858502758873,
+    "tau": 0.059256325887868444
+  },
+  "StackCube-v1": {
+    "base_success": 0.2727272727272727,
+    "base_utility": 0.7763947045261209,
+    "coverage": 0.3333333333333333,
+    "execute_generated": 0.3333333333333333,
+    "fallback_rate": 0.6666666666666666,
+    "hidden_chart_oracle_success": 1.0,
+    "hidden_chart_oracle_utility": 2.0,
+    "outcome_ptr": 0.6363636363636364,
+    "pairwise_causal_calibration_accuracy": 0.49359663297649975,
+    "pairwise_causal_calibration_confidence": 0.5413161148546212,
+    "pairwise_causal_calibration_ece": 0.10435404157033219,
+    "pairwise_causal_calibration_pairs": 67.45454545454545,
+    "predicted_margin": 0.14918837663928425,
+    "proposal_oracle_success": 0.5151515151515151,
+    "proposal_oracle_utility": 1.1567951707225856,
+    "selected_candidate_index": 5.393939393939394,
+    "selected_success": 0.30303030303030304,
+    "selected_success_gain_over_base": 0.030303030303030304,
+    "selected_utility": 0.7681677698185949,
+    "selected_utility_gain_over_base": -0.008226934707526005,
+    "selector_regret": 0.39222038604996423,
+    "success_selector_gap": 0.21212121212121213,
+    "success_support_gap": 0.48484848484848486,
+    "support_gap": 0.8432048292774142,
+    "tau": 0.22913040029030962
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/split_hash.txt

@@ -0,0 +1,14 @@
+{
+  "calibration_target_index": {
+    "content_hash": "67f7b4a692f7c7e71da24378dc71e3399c5d35afc14c494869eba6087b765c42",
+    "retrieval_index_allowed": true,
+    "split": "train",
+    "split_hash": "867a94a1c67686c082765671a7bf8d74a9d633085e1b2619c83197c60bf58372"
+  },
+  "eval_target_index": {
+    "content_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+    "retrieval_index_allowed": false,
+    "split": "test",
+    "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f"
+  }
+}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/table.tex

@@ -0,0 +1,8 @@
+% Auto-generated by scripts/eval_learned_dominance_selector.py
+\begin{tabular}{lrrrrrrrrr}
+\toprule
+Rows & Coverage & Fallback & Base succ. & Selected succ. & Oracle succ. & OutcomePTR & Succ. support gap & Succ. selector gap & Cal. ECE \\
+\midrule
+144 & 0.6458 & 0.3542 & 0.2917 & 0.3264 & 0.5694 & 0.5486 & 0.2014 & 0.2500 & 0.0188 \\
+\bottomrule
+\end{tabular}

workspace/runs/ctt_base_context_obs_learned_dominance_score_context_chartcompat_utility_task_envclip_k16_train_to_test/train.log

@@ -0,0 +1,5 @@
+trained ridge dominance calibrator on calibration measured rows only
+fit_objective=pointwise
+selected_lambda=100.0
+tau={"LiftPegUpright-v1": -0.085777, "PickCube-v1": 0.422526, "PullCube-v1": -0.174116, "PushCube-v1": 0.059256, "StackCube-v1": 0.22913, "__global__": -0.100999}
+calibration_input=runs/ctt_base_context_obs_train_cal_envclip_k16_rollout_comparison/combined_measured_candidates.json