auto-sync 2026-07-04T08:12:11Z workspace (part 5)

workspace/runs/ctt_selector_diagnostic_sweep/command.txt

@@ -1 +1 @@
-python scripts/build_selector_diagnostic_sweep.py 
+python scripts/build_selector_diagnostic_sweep.py --out-dir runs/ctt_selector_diagnostic_sweep

workspace/runs/ctt_selector_diagnostic_sweep/config.yaml

@@ -4,6 +4,15 @@ proposal_oracle_min: 0.5
 selector_gap_max: 0.03
 metrics:
   - runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_bundle_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_bundle_smoke_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_chart_bundle_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_chart_bundle_explore_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_explore_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_score_bundle_explore_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_envclip_k16_train_to_test/metrics.json
+  - runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_explore_envclip_k16_train_to_test/metrics.json
   - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json
   - runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json
   - runs/ctt_dominance_utility_energy_val_to_test_seed0/metrics.json

workspace/runs/ctt_selector_diagnostic_sweep/data_hash.txt

@@ -1 +1 @@
-fe282d92b43f35c62c14a40a1cce310487bf3bc1adbba59184d729da713d2129
+26fc5f6687ba0c88755ebf187c15e51702b8685a7339d484348df021668a3b7e

workspace/runs/ctt_selector_diagnostic_sweep/eval.log

@@ -1,3 +1,3 @@
-num_inputs=18
+num_inputs=27
 families=K8 per-dim trainmax,K8 tanh,K16 env_clip,K16 env_clip utility-energy
 overall_pass=False

workspace/runs/ctt_selector_diagnostic_sweep/metrics.json

@@ -85,7 +85,7 @@
       "success_support_gap": 0.2013888888888889
     }
   ],
-  "data_hash": "fe282d92b43f35c62c14a40a1cce310487bf3bc1adbba59184d729da713d2129",
+  "data_hash": "26fc5f6687ba0c88755ebf187c15e51702b8685a7339d484348df021668a3b7e",
   "gates": [
     {
       "family": "K8 per-dim trainmax",
@@ -126,6 +126,15 @@
   ],
   "input_metrics": [
     "runs/ctt_base_context_obs_learned_dominance_chartcompat_obs_utility_task_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_bundle_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_bundle_smoke_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_chart_bundle_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_chart_bundle_explore_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_explore_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_score_bundle_explore_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_envclip_k16_train_to_test/metrics.json",
+    "runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_explore_envclip_k16_train_to_test/metrics.json",
     "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test/metrics.json",
     "runs/ctt_base_context_obs_dominance_envclip_k16_train_to_test_tau0/metrics.json",
     "runs/ctt_dominance_utility_energy_val_to_test_seed0/metrics.json",
@@ -144,7 +153,7 @@
     "runs/ctt_base_context_obs_learned_dominance_context_tangent_success_perdim_trainmax_train_to_test/metrics.json",
     "runs/ctt_base_context_obs_dominance_perdim_trainmax_train_to_test/metrics.json"
   ],
-  "num_inputs": 18,
+  "num_inputs": 27,
   "overall_pass": false,
   "report_type": "ctt_selector_diagnostic_sweep",
   "rows": [
@@ -169,6 +178,195 @@
       "success_selector_gap": 0.24305555555555555,
       "success_support_gap": 0.2013888888888889
     },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.015103641044980439,
+      "coverage": 0.7708333333333334,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.22916666666666666,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_bundle_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.2708333333333333,
+      "selector": "bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.5607294605369033,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.3055555555555556,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.015103641044980439,
+      "coverage": 0.7708333333333334,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.22916666666666666,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_bundle_smoke_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.2708333333333333,
+      "selector": "bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.5607294605369033,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.3055555555555556,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.025456600726330422,
+      "coverage": 0.6111111111111112,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.3888888888888889,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_chart_bundle_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.3125,
+      "selector": "chart_bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.501692300305068,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2777777777777778,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.025456600726330422,
+      "coverage": 0.6111111111111112,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.3888888888888889,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_chart_bundle_explore_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.3125,
+      "selector": "chart_bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.501692300305068,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2777777777777778,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.06110488835310939,
+      "coverage": 0.5625,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.4375,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.2916666666666667,
+      "selector": "chart_source_bundle_consensus/success_weighted_margin, bonus=2.0",
+      "selector_regret": 0.5157032941917552,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2847222222222222,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.06110488835310939,
+      "coverage": 0.5625,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.4375,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_chart_source_bundle_explore_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.2916666666666667,
+      "selector": "chart_source_bundle_consensus/success_weighted_margin, bonus=2.0",
+      "selector_regret": 0.5157032941917552,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2847222222222222,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.013795704490294138,
+      "coverage": 0.7569444444444444,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.24305555555555555,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_score_bundle_explore_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.2638888888888889,
+      "selector": "score_bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.5747407733013259,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.3125,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.026027092934842826,
+      "coverage": 0.5694444444444444,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.4305555555555556,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.3125,
+      "selector": "score_chart_bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.4994775033636769,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2777777777777778,
+      "success_support_gap": 0.2013888888888889
+    },
+    {
+      "base_success": 0.2916666666666667,
+      "calibration_ece": 0.026027092934842826,
+      "coverage": 0.5694444444444444,
+      "data_hash": "0c38d73dc9bc4ab756e12006671b5d12f1a23358b281e54b3f6a9ae3cbf6e9af",
+      "fallback_rate": 0.4305555555555556,
+      "family": "K16 env_clip",
+      "hidden_chart_oracle_success": 0.7291666666666666,
+      "k": 16,
+      "outcome_ptr": 0.5486111111111112,
+      "proposal_oracle_success": 0.5694444444444444,
+      "report_type": "learned_dominance_selector_eval",
+      "run_path": "runs/ctt_base_context_obs_learned_dominance_score_chart_bundle_explore_envclip_k16_train_to_test",
+      "seed": "pooled",
+      "selected_success": 0.3125,
+      "selector": "score_chart_bundle_consensus/utility_margin, bonus=1.0",
+      "selector_regret": 0.4994775033636769,
+      "split_hash": "84151df9e88c219b76751677558cc5bbbc19082bf358cea7330158e9e403e25f",
+      "success_selector_gap": 0.2777777777777778,
+      "success_support_gap": 0.2013888888888889
+    },
     {
       "base_success": 0.2916666666666667,
       "calibration_ece": 0.16331725144301124,
@@ -529,7 +727,7 @@
   ],
   "schema_version": 1,
   "selection_rule": "best selected_success per diagnostic family; all candidate rows retained",
-  "split_hash": "08acb82a67e5b95e3e8339cf57cfcb2695aca7e8f0b794889e754059bdd0b6b6",
+  "split_hash": "6d97fc8f919974eb75db4c325e8d3497cb40806073eedf478d2d7655b39ca981",
   "thresholds": {
     "proposal_oracle_min": 0.5,
     "selected_min": 0.4745,

workspace/runs/ctt_selector_diagnostic_sweep/metrics_by_seed.json

@@ -30,13 +30,13 @@
     "success_support_gap": 0.2013888888888889
   },
   "pooled": {
-    "base_success": 0.325,
-    "calibration_ece": 0.055655260342575806,
-    "coverage": 0.22916666666666669,
-    "outcome_ptr": 0.42083333333333317,
-    "proposal_oracle_success": 0.3555555555555557,
-    "selected_success": 0.3282407407407407,
-    "success_selector_gap": 0.13981481481481478,
-    "success_support_gap": 0.3930555555555556
+    "base_success": 0.31249999999999983,
+    "calibration_ece": 0.046000377322810725,
+    "coverage": 0.38425925925925924,
+    "outcome_ptr": 0.46874999999999983,
+    "proposal_oracle_success": 0.4357638888888888,
+    "selected_success": 0.3151041666666667,
+    "success_selector_gap": 0.19589120370370372,
+    "success_support_gap": 0.3211805555555556
   }
 }

workspace/runs/ctt_selector_diagnostic_sweep/metrics_by_task.json

@@ -1,13 +1,13 @@
 {
   "K16 env_clip": {
-    "base_success": 0.2916666666666667,
-    "calibration_ece": 0.11388905660618416,
-    "coverage": 0.3009259259259259,
-    "outcome_ptr": 0.5486111111111112,
-    "proposal_oracle_success": 0.5694444444444444,
-    "selected_success": 0.3078703703703704,
-    "success_selector_gap": 0.28009259259259256,
-    "success_support_gap": 0.20138888888888892
+    "base_success": 0.29166666666666663,
+    "calibration_ece": 0.050903943368947734,
+    "coverage": 0.5572916666666666,
+    "outcome_ptr": 0.548611111111111,
+    "proposal_oracle_success": 0.5694444444444445,
+    "selected_success": 0.29687499999999994,
+    "success_selector_gap": 0.28703703703703703,
+    "success_support_gap": 0.20138888888888887
   },
   "K16 env_clip utility-energy": {
     "base_success": 0.2916666666666667,

workspace/runs/ctt_selector_diagnostic_sweep/split_hash.txt

@@ -1 +1 @@
-08acb82a67e5b95e3e8339cf57cfcb2695aca7e8f0b794889e754059bdd0b6b6
+6d97fc8f919974eb75db4c325e8d3497cb40806073eedf478d2d7655b39ca981