Auto-sync: 2026-06-29 09:44:12

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_consensus_grid035040045_noopbonus003",
-  "generated_utc": "2026-06-29T12:01:00+00:00",
+  "generated_utc": "2026-06-29T13:46:02+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.07130434782608691,
     "best_clean_vs_h16": 0.05681159420289855,
@@ -578,6 +578,272 @@
       "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_safe_k4_grid025035050_margin0p20_summary.json",
       "std_success": 0.012541047914657353
     },
+    "residual_k4_compose_grid035040045": {
+      "ci95_success": 0.038402227815547745,
+      "label": "K4 composed type-consensus tangents, scales 0.35/0.40/0.45",
+      "mean_action_mse_to_best": 0.48167677350261284,
+      "mean_progress": 0.5596326749873518,
+      "mean_success": 0.34086956521739137,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.22953096331437725,
+          "std_success": 0.09991475281512771
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.31274633176807093,
+          "std_success": 0.024772567569059156
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20474165211007314,
+          "std_success": 0.019188581862611215
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7538575019089029,
+          "std_success": 0.08379254382407199
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.18023828023828026,
+          "std_success": 0.054059397468562184
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.5628787715733051,
+        "1": 0.4294346837553641,
+        "2": 0.4527168651791694
+      },
+      "seed_progress": {
+        "0": 0.5319681812231631,
+        "1": 0.5689693509657746,
+        "2": 0.5779604927731176
+      },
+      "seed_success": {
+        "0": 0.3234782608695652,
+        "1": 0.34608695652173915,
+        "2": 0.35304347826086957
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1531,
+        "retrieval_residual_residual_near_miss": 5,
+        "retrieval_residual_residual_near_miss+residual_no_op": 3,
+        "retrieval_residual_residual_near_miss+residual_wrong_direction": 1,
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": 4,
+        "retrieval_residual_residual_no_op": 18,
+        "retrieval_residual_residual_no_op+residual_random_negative": 125,
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": 6,
+        "retrieval_residual_residual_wrong_direction+residual_random_negative": 1,
+        "retrieval_residual_residual_wrong_direction+residual_wrong_gripper": 2,
+        "retrieval_residual_residual_wrong_gripper": 18,
+        "retrieval_residual_residual_wrong_gripper+residual_random_negative": 11
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1562,
+        "0.4": 25,
+        "0.45": 138
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1531.0,
+          "mean_progress": 0.5599307770840422,
+          "success_rate": 0.3514043109079033
+        },
+        "retrieval_residual_residual_near_miss": {
+          "count": 5.0,
+          "mean_progress": 0.8016590535640716,
+          "success_rate": 0.6
+        },
+        "retrieval_residual_residual_near_miss+residual_no_op": {
+          "count": 3.0,
+          "mean_progress": 0.5828627347946167,
+          "success_rate": 0.0
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_direction": {
+          "count": 1.0,
+          "mean_progress": 1.0,
+          "success_rate": 1.0
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": {
+          "count": 4.0,
+          "mean_progress": 0.6947316378355026,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 18.0,
+          "mean_progress": 0.7534612036413617,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_no_op+residual_random_negative": {
+          "count": 125.0,
+          "mean_progress": 0.46661006832122803,
+          "success_rate": 0.128
+        },
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": {
+          "count": 6.0,
+          "mean_progress": 0.9279975295066833,
+          "success_rate": 0.8333333333333334
+        },
+        "retrieval_residual_residual_wrong_direction+residual_random_negative": {
+          "count": 1.0,
+          "mean_progress": 0.8137040138244629,
+          "success_rate": 0.0
+        },
+        "retrieval_residual_residual_wrong_direction+residual_wrong_gripper": {
+          "count": 2.0,
+          "mean_progress": 0.8591917157173157,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 18.0,
+          "mean_progress": 0.659929613272349,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_wrong_gripper+residual_random_negative": {
+          "count": 11.0,
+          "mean_progress": 0.6099244315515865,
+          "success_rate": 0.36363636363636365
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_compose_grid035040045_safe_margin0p20_summary.json",
+      "std_success": 0.015457729421418421
+    },
+    "residual_k4_compose_grid035040045_noopbonus003": {
+      "ci95_success": 0.03920403094295297,
+      "label": "K4 composed type-consensus tangents, scales 0.35/0.40/0.45, no-op bonus 0.03",
+      "mean_action_mse_to_best": 0.48164921497764146,
+      "mean_progress": 0.5600020351956472,
+      "mean_success": 0.34144927536231884,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.22953096331437725,
+          "std_success": 0.09991475281512771
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.31455792597096943,
+          "std_success": 0.023416763440172416
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20474165211007314,
+          "std_success": 0.019188581862611215
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7538575019089029,
+          "std_success": 0.08379254382407199
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.18023828023828026,
+          "std_success": 0.054059397468562184
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.563375490748364,
+        "1": 0.4286549947750957,
+        "2": 0.4529171594094647
+      },
+      "seed_progress": {
+        "0": 0.5323185247628262,
+        "1": 0.5697598444961983,
+        "2": 0.577927736327917
+      },
+      "seed_success": {
+        "0": 0.3234782608695652,
+        "1": 0.34782608695652173,
+        "2": 0.35304347826086957
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1521,
+        "retrieval_residual_residual_near_miss": 5,
+        "retrieval_residual_residual_near_miss+residual_no_op": 1,
+        "retrieval_residual_residual_near_miss+residual_wrong_direction": 1,
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": 4,
+        "retrieval_residual_residual_no_op": 35,
+        "retrieval_residual_residual_no_op+residual_random_negative": 122,
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": 5,
+        "retrieval_residual_residual_wrong_direction+residual_random_negative": 1,
+        "retrieval_residual_residual_wrong_direction+residual_wrong_gripper": 2,
+        "retrieval_residual_residual_wrong_gripper": 18,
+        "retrieval_residual_residual_wrong_gripper+residual_random_negative": 10
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1551,
+        "0.4": 27,
+        "0.45": 147
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1521.0,
+          "mean_progress": 0.5611518402145079,
+          "success_rate": 0.3530571992110454
+        },
+        "retrieval_residual_residual_near_miss": {
+          "count": 5.0,
+          "mean_progress": 0.8016590535640716,
+          "success_rate": 0.6
+        },
+        "retrieval_residual_residual_near_miss+residual_no_op": {
+          "count": 1.0,
+          "mean_progress": 0.6861283183097839,
+          "success_rate": 0.0
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_direction": {
+          "count": 1.0,
+          "mean_progress": 1.0,
+          "success_rate": 1.0
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": {
+          "count": 4.0,
+          "mean_progress": 0.6947316378355026,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 35.0,
+          "mean_progress": 0.648064909396427,
+          "success_rate": 0.37142857142857144
+        },
+        "retrieval_residual_residual_no_op+residual_random_negative": {
+          "count": 122.0,
+          "mean_progress": 0.46078647990695765,
+          "success_rate": 0.12295081967213115
+        },
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": {
+          "count": 5.0,
+          "mean_progress": 1.0,
+          "success_rate": 1.0
+        },
+        "retrieval_residual_residual_wrong_direction+residual_random_negative": {
+          "count": 1.0,
+          "mean_progress": 0.8137040138244629,
+          "success_rate": 0.0
+        },
+        "retrieval_residual_residual_wrong_direction+residual_wrong_gripper": {
+          "count": 2.0,
+          "mean_progress": 0.8591917157173157,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 18.0,
+          "mean_progress": 0.659929613272349,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_wrong_gripper+residual_random_negative": {
+          "count": 10.0,
+          "mean_progress": 0.5709168747067451,
+          "success_rate": 0.3
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_compose_grid035040045_safe_margin0p20_noopbonus0p03_summary.json",
+      "std_success": 0.01578047256674342
+    },
     "residual_k4_consensus": {
       "ci95_success": 0.04490086956521744,
       "label": "K4 mean-by-type tangent consensus",

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-29T12:01:00+00:00`
+Generated: `2026-06-29T13:46:02+00:00`
 
 ## Main Seed Statistics
 
@@ -43,6 +43,8 @@ Generated: `2026-06-29T12:01:00+00:00`
 | residual_k4_consensus_grid035040045_noopbonus003_consensus002 | K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03, consensus penalty 0.02 | 3 | 35.36% +/- 1.16 | +/- 2.88 | 56.82% | 0.398 | +5.62 pp |
 | residual_k4_consensus_grid035040045_noopbonus003_consensus005 | K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03, consensus penalty 0.05 | 3 | 35.36% +/- 1.16 | +/- 2.88 | 56.78% | 0.398 | +5.62 pp |
 | residual_k4_consensus_grid035040045_noopbonus003_consensus010 | K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03, consensus penalty 0.10 | 3 | 35.36% +/- 1.16 | +/- 2.88 | 56.75% | 0.397 | +5.62 pp |
+| residual_k4_compose_grid035040045 | K4 composed type-consensus tangents, scales 0.35/0.40/0.45 | 3 | 34.09% +/- 1.55 | +/- 3.84 | 55.96% | 0.482 | +4.35 pp |
+| residual_k4_compose_grid035040045_noopbonus003 | K4 composed type-consensus tangents, scales 0.35/0.40/0.45, no-op bonus 0.03 | 3 | 34.14% +/- 1.58 | +/- 3.92 | 56.00% | 0.482 | +4.41 pp |
 | repair_nearmiss_k4_grid025035050_margin020 | K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20 | 3 | 34.32% +/- 1.35 | +/- 3.36 | 55.97% | 0.394 | +4.58 pp |
 | repair_nearmiss_k4_grid035050075_margin020 | K4 near-miss-to-expert repair tangent, scales 0.35/0.50/0.75, margin 0.20 | 3 | 34.38% +/- 1.50 | +/- 3.73 | 56.05% | 0.394 | +4.64 pp |
 | repair_nearmiss_k4_grid025035050_margin010 | K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.10 | 3 | 34.14% +/- 1.48 | +/- 3.67 | 56.01% | 0.393 | +4.41 pp |

results/paper_table_status.json

@@ -1220,6 +1220,44 @@
       "best_config": null,
       "gain_vs_h16_policy": 0.05623188405797103
     },
+    {
+      "key": "retrieval_residual_k4_compose_grid035040045",
+      "label": "K4 composed type-consensus residual retrieval, scales 0.35/0.40/0.45, margin 0.20",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_compose_grid035040045_safe_margin0p20_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "local tangent composition without typed priors",
+      "fallback_success": null,
+      "pending_job": "14905956/14905963",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.3408695652173913,
+      "std_success": 0.015457729421418423,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.043478260869565244
+    },
+    {
+      "key": "retrieval_residual_k4_compose_grid035040045_noopbonus003",
+      "label": "K4 composed type-consensus residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_compose_grid035040045_safe_margin0p20_noopbonus0p03_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "local tangent composition on the current best typed prior",
+      "fallback_success": null,
+      "pending_job": "14905957/14905964",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.34144927536231884,
+      "std_success": 0.01578047256674342,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.04405797101449277
+    },
     {
       "key": "retrieval_repair_nearmiss_k4_grid025035050_margin020",
       "label": "K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20",

results/paper_table_status.md

@@ -67,6 +67,8 @@ Baseline h=16 policy: 29.74%
 | retrieval_residual_k4_mean_grid035040045_noopbonus003_consensus002 | K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03, consensus penalty 0.02 | complete | 35.36% | +5.62 pp | yes | no | no | train-neighbor tangent-consensus confidence on the current best typed prior |
 | retrieval_residual_k4_mean_grid035040045_noopbonus003_consensus005 | K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03, consensus penalty 0.05 | complete | 35.36% | +5.62 pp | yes | no | no | train-neighbor tangent-consensus confidence on the current best typed prior |
 | retrieval_residual_k4_mean_grid035040045_noopbonus003_consensus010 | K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03, consensus penalty 0.10 | complete | 35.36% | +5.62 pp | yes | no | no | train-neighbor tangent-consensus confidence on the current best typed prior |
+| retrieval_residual_k4_compose_grid035040045 | K4 composed type-consensus residual retrieval, scales 0.35/0.40/0.45, margin 0.20 | complete | 34.09% | +4.35 pp | yes | no | no | local tangent composition without typed priors |
+| retrieval_residual_k4_compose_grid035040045_noopbonus003 | K4 composed type-consensus residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03 | complete | 34.14% | +4.41 pp | yes | no | no | local tangent composition on the current best typed prior |
 | retrieval_repair_nearmiss_k4_grid025035050_margin020 | K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20 | complete | 34.32% | +4.58 pp | yes | no | no | deployment-clean corrective tangent transport from train near-misses back toward expert actions |
 | retrieval_repair_nearmiss_k4_grid035050075_margin020 | K4 near-miss-to-expert repair tangent, scales 0.35/0.50/0.75, margin 0.20 | complete | 34.38% | +4.64 pp | yes | no | no | repair-tangent scale diagnostic for near-miss counterfactual geometry |
 | retrieval_repair_nearmiss_k4_grid025035050_margin010 | K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.10 | complete | 34.14% | +4.41 pp | yes | no | no | repair-tangent abstention diagnostic for near-miss counterfactual geometry |