Auto-sync: 2026-06-29 17:18:25 (part 3)

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003",
-  "generated_utc": "2026-06-29T21:12:22+00:00",
+  "generated_utc": "2026-06-29T21:44:25+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.07304347826086954,
     "best_clean_vs_h16": 0.058550724637681184,
@@ -948,9 +948,101 @@
       "std_success": 0.011579701075616346
     },
     "residual_k4_composemasked_dropnmnoop_grid035040045": {
+      "ci95_success": 0.03115620450679224,
       "label": "K4 composed type-consensus tangents, masked, drop near-miss+no-op composite",
-      "missing": true,
-      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_summary.json"
+      "mean_action_mse_to_best": 0.40620721947428756,
+      "mean_progress": 0.5700427223033847,
+      "mean_success": 0.3547826086956522,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.26340865087329823,
+          "std_success": 0.06057881385012102
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.32419372093285137,
+          "std_success": 0.03154715976310507
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20844535581377688,
+          "std_success": 0.024567915948941907
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.757403601199683,
+          "std_success": 0.07781968814460603
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.19707329707329704,
+          "std_success": 0.0831467196651622
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3921792995427614,
+        "1": 0.3989143911279414,
+        "2": 0.42752796775215995
+      },
+      "seed_progress": {
+        "0": 0.5565006508119404,
+        "1": 0.5698966576515333,
+        "2": 0.5837308584466803
+      },
+      "seed_success": {
+        "0": 0.35130434782608694,
+        "1": 0.3443478260869565,
+        "2": 0.36869565217391304
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1628,
+        "retrieval_residual_residual_near_miss": 8,
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": 5,
+        "retrieval_residual_residual_no_op": 38,
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": 27,
+        "retrieval_residual_residual_wrong_gripper": 19
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1643,
+        "0.4": 10,
+        "0.45": 72
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1628.0,
+          "mean_progress": 0.5603744033500014,
+          "success_rate": 0.34459459459459457
+        },
+        "retrieval_residual_residual_near_miss": {
+          "count": 8.0,
+          "mean_progress": 0.8760369084775448,
+          "success_rate": 0.75
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": {
+          "count": 5.0,
+          "mean_progress": 0.7557853102684021,
+          "success_rate": 0.6
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 38.0,
+          "mean_progress": 0.7757954887653652,
+          "success_rate": 0.5263157894736842
+        },
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": {
+          "count": 27.0,
+          "mean_progress": 0.6916872307106301,
+          "success_rate": 0.48148148148148145
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 19.0,
+          "mean_progress": 0.6363769314791027,
+          "success_rate": 0.47368421052631576
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_summary.json",
+      "std_success": 0.012541047914657353
     },
     "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003": {
       "ci95_success": 0.024567904418045634,
@@ -1049,6 +1141,200 @@
       "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_summary.json",
       "std_success": 0.009889114266221456
     },
+    "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001": {
+      "ci95_success": 0.024567904418045634,
+      "label": "K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.01",
+      "mean_action_mse_to_best": 0.406297814965626,
+      "mean_progress": 0.571048675666919,
+      "mean_success": 0.35594202898550725,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.26340865087329823,
+          "std_success": 0.06057881385012102
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.3278169093386485,
+          "std_success": 0.02533117351097172
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20474165211007314,
+          "std_success": 0.019188581862611215
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.757403601199683,
+          "std_success": 0.07781968814460603
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.20044030044030045,
+          "std_success": 0.08896982358351117
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.39301921862784933,
+        "1": 0.39824521353549286,
+        "2": 0.42762901273353593
+      },
+      "seed_progress": {
+        "0": 0.5583201415569562,
+        "1": 0.5714951352645522,
+        "2": 0.5833307501792486
+      },
+      "seed_success": {
+        "0": 0.35304347826086957,
+        "1": 0.34782608695652173,
+        "2": 0.36695652173913046
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1613,
+        "retrieval_residual_residual_near_miss": 9,
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": 4,
+        "retrieval_residual_residual_no_op": 56,
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": 24,
+        "retrieval_residual_residual_wrong_gripper": 19
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1631,
+        "0.4": 11,
+        "0.45": 83
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1613.0,
+          "mean_progress": 0.560971513695882,
+          "success_rate": 0.3453192808431494
+        },
+        "retrieval_residual_residual_near_miss": {
+          "count": 9.0,
+          "mean_progress": 0.8691005806128184,
+          "success_rate": 0.6666666666666666
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": {
+          "count": 4.0,
+          "mean_progress": 0.898208275437355,
+          "success_rate": 0.75
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 56.0,
+          "mean_progress": 0.7249722117225507,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": {
+          "count": 24.0,
+          "mean_progress": 0.671148752172788,
+          "success_rate": 0.4583333333333333
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 19.0,
+          "mean_progress": 0.6363769314791027,
+          "success_rate": 0.47368421052631576
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p01_summary.json",
+      "std_success": 0.009889114266221456
+    },
+    "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002": {
+      "ci95_success": 0.024567904418045634,
+      "label": "K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.02",
+      "mean_action_mse_to_best": 0.4063192498065747,
+      "mean_progress": 0.571048675666919,
+      "mean_success": 0.35594202898550725,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.26340865087329823,
+          "std_success": 0.06057881385012102
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.3278169093386485,
+          "std_success": 0.02533117351097172
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20474165211007314,
+          "std_success": 0.019188581862611215
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.757403601199683,
+          "std_success": 0.07781968814460603
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.20044030044030045,
+          "std_success": 0.08896982358351117
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3930835231506954,
+        "1": 0.39824521353549286,
+        "2": 0.42762901273353593
+      },
+      "seed_progress": {
+        "0": 0.5583201415569562,
+        "1": 0.5714951352645522,
+        "2": 0.5833307501792486
+      },
+      "seed_success": {
+        "0": 0.35304347826086957,
+        "1": 0.34782608695652173,
+        "2": 0.36695652173913046
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1612,
+        "retrieval_residual_residual_near_miss": 10,
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": 4,
+        "retrieval_residual_residual_no_op": 56,
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": 24,
+        "retrieval_residual_residual_wrong_gripper": 19
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1630,
+        "0.4": 11,
+        "0.45": 84
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1612.0,
+          "mean_progress": 0.5606991635182741,
+          "success_rate": 0.34491315136476425
+        },
+        "retrieval_residual_residual_near_miss": {
+          "count": 10.0,
+          "mean_progress": 0.8821905225515365,
+          "success_rate": 0.7
+        },
+        "retrieval_residual_residual_near_miss+residual_wrong_gripper": {
+          "count": 4.0,
+          "mean_progress": 0.898208275437355,
+          "success_rate": 0.75
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 56.0,
+          "mean_progress": 0.7249722117225507,
+          "success_rate": 0.5
+        },
+        "retrieval_residual_residual_no_op+residual_wrong_gripper": {
+          "count": 24.0,
+          "mean_progress": 0.671148752172788,
+          "success_rate": 0.4583333333333333
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 19.0,
+          "mean_progress": 0.6363769314791027,
+          "success_rate": 0.47368421052631576
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p02_summary.json",
+      "std_success": 0.009889114266221456
+    },
     "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_srcscorebonus002": {
       "ci95_success": 0.02131295542962397,
       "label": "K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, no-op bonus 0.03, source-score bonus 0.02",

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-29T21:12:22+00:00`
+Generated: `2026-06-29T21:44:25+00:00`
 
 ## Main Seed Statistics
 
@@ -47,8 +47,10 @@ Generated: `2026-06-29T21:12:22+00:00`
 | 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 |
 | residual_k4_composemasked_grid035040045 | K4 composed type-consensus tangents, masked, scales 0.35/0.40/0.45 | 3 | 35.30% +/- 1.22 | +/- 3.02 | 56.91% | 0.410 | +5.57 pp |
 | residual_k4_composemasked_grid035040045_noopbonus003 | K4 composed type-consensus tangents, masked, scales 0.35/0.40/0.45, no-op bonus 0.03 | 3 | 35.54% +/- 1.02 | +/- 2.53 | 57.02% | 0.411 | +5.80 pp |
-| residual_k4_composemasked_dropnmnoop_grid035040045 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite | 0 | missing | missing | missing | missing | missing |
+| residual_k4_composemasked_dropnmnoop_grid035040045 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite | 3 | 35.48% +/- 1.25 | +/- 3.12 | 57.00% | 0.406 | +5.74 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite | 3 | 35.59% +/- 0.99 | +/- 2.46 | 57.07% | 0.406 | +5.86 pp |
+| residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001 | K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.01 | 3 | 35.59% +/- 0.99 | +/- 2.46 | 57.10% | 0.406 | +5.86 pp |
+| residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002 | K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.02 | 3 | 35.59% +/- 0.99 | +/- 2.46 | 57.10% | 0.406 | +5.86 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, source-score bonus 0.02 | 3 | 35.48% +/- 1.22 | +/- 3.02 | 57.02% | 0.408 | +5.74 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_srcscorebonus002 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, no-op bonus 0.03, source-score bonus 0.02 | 3 | 35.54% +/- 0.86 | +/- 2.13 | 57.04% | 0.408 | +5.80 pp |
 | residual_k4_composemasked_dropnmnoop_l2comp002_grid035040045_noopbonus003 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, composite L2 penalty 0.02 | 3 | 35.54% +/- 1.06 | +/- 2.64 | 57.06% | 0.405 | +5.80 pp |

results/paper_core_results.md

@@ -55,8 +55,11 @@ lattice is `+27.25 pp`, and the remaining clean-to-same-state proposal gap is
 | K4 mean-by-type residual retrieval + train-neighbor consensus penalty | No | No | 35.19-35.36% | +5.45-5.62 pp | Penalizing high-dispersion local tangent families is coherent but over-abstains by one success; geometric confidence does not improve the sparse no-op scale-grid row |
 | K4 composed type-consensus residual retrieval, masked | No | No | 35.30% | +5.57 pp | Clean anti-goal composite masking removes the confound; tangent composition alone near-ties but does not beat mean-by-type scale-grid transport |
 | K4 composed type-consensus residual retrieval, masked + no-op prior 0.03 | No | No | 35.54% | +5.80 pp | Previous clean best; masked local tangent composition adds one success over the scale-grid no-op row while staying sparse |
-| K4 composed type-consensus residual retrieval, masked + exact near-miss/no-op incompatibility mask | No | No | 35.59% | +5.86 pp | Current best clean diagnostic; dropping only the weak `near_miss+no_op` composite preserves useful singleton/composite tangents and lowers action MSE |
+| K4 composed type-consensus residual retrieval, exact compatibility mask without typed prior | No | No | 35.48% | +5.74 pp | Isolates the algebraic compatibility mask: it improves pure masked composition but does not fully replace the sparse typed no-op prior |
+| K4 composed type-consensus residual retrieval, no-op prior + exact near-miss/no-op incompatibility mask | No | No | 35.59% | +5.86 pp | Current best clean diagnostic; dropping only the weak `near_miss+no_op` composite preserves useful singleton/composite tangents and lowers action MSE |
+| K4 composed type-consensus residual retrieval, exact compatibility mask + no-op/near-miss singleton priors | No | No | 35.59% | +5.86 pp | Exact singleton near-miss priors 0.01/0.02 tie the top row and raise progress slightly; transferred near-miss tangents are high-precision but too sparse to close the proposal gap |
 | K4 composed type-consensus residual retrieval, exact compatibility mask + source-score prior | No | No | 35.48% | +5.74 pp | Measured train-source reward confidence does not replace the sparse typed prior on the compatible chart; useful negative calibration |
+| K4 composed type-consensus residual retrieval, exact compatibility mask + no-op/source-score priors | No | No | 35.54% | +5.80 pp | Adding measured train-source reward confidence to the current typed-prior chart ties near-best but does not beat exact compatibility masking alone |
 | K4 composed type-consensus residual retrieval, masked + component-wise no-op prior | No | No | 35.36% | +5.62 pp | Propagating the no-op prior into every composite over-selects composed tangents and drops below the exact-prior row; useful negative calibration |
 | K4 composed type-consensus residual retrieval, masked + composite trust penalty | No | No | 35.48-35.54% | +5.74-5.80 pp | Composite-only L2 lowers action MSE but does not improve success; combined with the exact compatibility mask it reaches 35.54% with the lowest MSE, one seed below the 35.59% top row |
 | K4 repair-tangent residual transport | No | No | 34.14-34.43% | +4.41-4.70 pp | Reversing residuals into failure-to-expert repair tangents is a clean negative diagnostic; the current gain is not recovered by transporting near-miss-to-expert vectors |
@@ -106,7 +109,7 @@ Suggested main-table rows:
 13. K4 mean-by-type residual retrieval + fixed-scale no-op prior plateau, canonical 0.03
 14. K4 mean-by-type residual retrieval + scale-grid no-op prior 0.03
 15. K4 masked composed type-consensus residual retrieval + no-op prior 0.03
-16. K4 masked composed type-consensus residual retrieval + exact near-miss/no-op compatibility mask
+16. K4 masked composed type-consensus residual retrieval + exact near-miss/no-op compatibility mask, with and without typed no-op prior
 17. K4 masked composed type-consensus residual retrieval + source-score prior on the exact compatibility chart
 18. K4 masked composed type-consensus residual retrieval + component-wise no-op prior diagnostic
 19. K4 masked composed type-consensus residual retrieval + composite trust penalty diagnostic

results/paper_story_memo.md

@@ -29,8 +29,9 @@ when queried on proposal geometry that matches those local counterfactuals.
 | Clean residual transport behaves like sparse intervention | the best clean row abstains to zero-residual policy on 93.5% of states, while selected nonzero residuals succeed at 50.9% vs 34.5% for abstention | Stronger clean-mechanism framing |
 | Tangent consensus is close but needs sparse typing | K4 mean-by-type residual consensus reaches 34.96%; a small no-op residual prior plateau at 0.025-0.035 raises fixed-scale transport to 35.25% | Fixed-scale clean diagnostic |
 | Field-gated tangent length calibration improves the clean bridge | K4 mean-by-type scale grid 0.35/0.40/0.45 with no-op bonus 0.03 reaches 35.42%; the source-score version reaches 35.30%. Upper 0.40/0.45/0.50 nearly ties at 35.36%, while wide 0.35/0.45/0.55 drops to 35.13% | Previous clean best; local scale calibration, not a larger-step effect |
-| Masked tangent composition gives a small clean lift | K4 composed type-consensus transport is 35.30% after anti-goal composite masking, reaches 35.54% with the typed no-op prior, and reaches 35.59% when only `near_miss+no_op` is additionally masked | Current best clean result; controlled compatible tangent-chart composition, not unmasked proposal accumulation |
-| Train-source score priors do not explain the compatible tangent-chart gain | replacing the typed no-op prior with a measured train-source reward-score prior on the exact compatibility chart reaches 35.48%, below the 35.59% top row | Negative diagnostic: compatibility masking matters more than continuous train-source confidence on this chart |
+| Masked tangent composition gives a small clean lift | K4 composed type-consensus transport is 35.30% after anti-goal composite masking; exact `near_miss+no_op` masking alone reaches 35.48%; typed no-op prior reaches 35.54%; combining the exact compatibility mask with the typed prior reaches 35.59% | Current best clean result; controlled compatible tangent-chart composition, not unmasked proposal accumulation |
+| Train-source score priors do not explain the compatible tangent-chart gain | replacing the typed no-op prior with a measured train-source reward-score prior on the exact compatibility chart reaches 35.48%; adding it to the typed no-op row reaches 35.54%, still below the 35.59% top row | Negative diagnostic: compatibility masking and sparse typed abstention matter more than continuous train-source confidence on this chart |
+| Singleton near-miss priors do not open a new clean route | adding a small exact `residual_near_miss` singleton prior 0.01 or 0.02 on the compatible chart ties the 35.59% top row and slightly raises progress to 57.10% | Tie diagnostic: revived near-miss tangents are high-precision but too sparse to close the clean-to-same-state proposal gap |
 | Component-wise composite priors do not add the gain | propagating the no-op bonus into composite types reaches 35.36%, below the exact-prior masked composition row at 35.54% and the exact compatibility row at 35.59% | Negative/near-tie diagnostic: sparse exact priors are better than broadly rewarding every no-op-containing composite |
 | Composite trust-radius penalty explains but does not improve the success top line | composite-only L2 penalty 0.02 ties 35.54% while lowering action MSE from 0.4106 to 0.4079; with the exact compatibility mask it lowers MSE further to 0.4048 but drops back to 35.54% | Tie/negative diagnostic: composed tangents need a local trust radius, but compatibility masking gives the success optimum |
 | Minimum-energy residual regularization does not add the gain | action L2 penalty 0.05 ties the previous 35.42% scale-grid row, while 0.10/0.20 reach 35.36% | Negative/tie diagnostic: the clean bridge is not explained by shortest-action bias |
@@ -78,7 +79,7 @@ clean proposal result, the intended main rows are:
 14. K4 mean-by-type tangent consensus: 34.96%
 15. K4 mean-by-type tangent consensus + typed no-op prior 0.025-0.035: 35.25%
 16. K4 mean-by-type tangent consensus + scale-grid typed no-op prior: 35.42%
-17. K4 masked composed type-consensus tangent transport: 35.30%; with typed no-op prior: 35.54%; with exact near-miss/no-op compatibility mask: 35.59%; with source-score prior on that exact chart: 35.48%
+17. K4 masked composed type-consensus tangent transport: 35.30%; with exact near-miss/no-op compatibility mask only: 35.48%; with typed no-op prior: 35.54%; with exact compatibility mask + typed no-op prior: 35.59%; adding singleton near-miss prior 0.01/0.02 ties 35.59%; with source-score prior on that exact chart: 35.48%; with typed no-op + source-score: 35.54%
 18. K4 masked composed type-consensus tangent transport + component-wise no-op prior: 35.36%
 19. K4 masked composed type-consensus tangent transport + composite L2 trust penalty: 35.54% at 0.02; 35.48% at 0.05; 35.54% with exact compatibility mask + L2 0.02
 20. K4 mean-by-type tangent consensus + upper/wide scale diagnostics: 35.36% for 0.40/0.45/0.50; 35.13% for 0.35/0.45/0.55
@@ -135,10 +136,11 @@ test-time search. The cleaner novelty is:
 
 ## Job Status
 
-Last checked: `2026-06-29 21:09 UTC`. The K4 masked composed type-consensus
+Last checked: `2026-06-29 21:42 UTC`. The K4 masked composed type-consensus
 sweep completed and produced a new clean best, 35.59%, when the exact
 `residual_near_miss+residual_no_op` composite is additionally masked. Pure
-masked composition reached 35.30%, masked composition with the exact typed
+masked composition reached 35.30%, exact compatibility masking without typed
+or source-score priors reached 35.48%, masked composition with the exact typed
 no-op prior reached 35.54%, and raw selected candidate types show no
 random-negative or wrong-direction composite leak. The paper table and paired
 analysis now use the exact compatibility-mask row as `best_clean_key`. The
@@ -147,11 +149,12 @@ no-op prior row. Composite-only L2 trust penalties reduce action MSE but do not
 improve success: standalone L2 0.02 ties 35.54%, L2 0.05 reaches 35.48%, and
 exact compatibility + L2 0.02 reaches 35.54% with the lowest MSE.
 On the same exact compatibility chart, a train-measured source-score prior
-without the typed no-op prior completed at 35.48%, below the 35.59% top row.
-The no-op+source-score calibration row is still running as `14926714/14926726`.
-A clean mask-only ablation, exact `near_miss+no_op` drop without typed no-op or
-source-score priors, is queued as `14927919/14927921`; this isolates the
-compatibility mask from prior tuning.
+without the typed no-op prior completed at 35.48%, below the 35.59% top row;
+adding source-score to the typed no-op exact-mask row completed at 35.54%, also
+below the top row.
+The singleton near-miss revival diagnostic completed: exact `residual_near_miss`
+bonuses 0.01 and 0.02 both tie the 35.59% top row and slightly raise mean
+progress to 57.10%, but do not improve success.
 
 - `14858328`-`14858333`: completed train-split `field_selected_noexpert_bc5`;
   direct rollout is 26.84%, field-guided best is 27.65%.

results/paper_table_status.json

@@ -1306,14 +1306,14 @@
       "story_role": "exact compatibility mask without typed no-op prior",
       "fallback_success": null,
       "pending_job": "14927919/14927921",
-      "path_exists": false,
-      "status": "pending",
-      "success": null,
-      "std_success": null,
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35478260869565215,
+      "std_success": 0.012541047914657353,
       "completed_seeds": null,
-      "num_completed": null,
+      "num_completed": 3,
       "best_config": null,
-      "gain_vs_h16_policy": null
+      "gain_vs_h16_policy": 0.05739130434782608
     },
     {
       "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003",
@@ -1334,6 +1334,44 @@
       "best_config": null,
       "gain_vs_h16_policy": 0.058550724637681184
     },
+    {
+      "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
+      "label": "K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p01_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "revive high-precision singleton near-miss tangents without boosting toxic near-miss+no-op composites",
+      "fallback_success": null,
+      "pending_job": "14929087/14929152",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35594202898550725,
+      "std_success": 0.009889114266221456,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.058550724637681184
+    },
+    {
+      "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002",
+      "label": "K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.02",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p02_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "stronger singleton near-miss revival prior on the compatible local tangent chart",
+      "fallback_success": null,
+      "pending_job": "14929120/14929175",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35594202898550725,
+      "std_success": 0.009889114266221456,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.058550724637681184
+    },
     {
       "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002",
       "label": "K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, source-score bonus 0.02",
@@ -2541,6 +2579,7 @@
     "Train-state counterfactual residual retrieval is a positive clean bridge but remains below the current clean best.",
     "KNN counterfactual residual retrieval is a positive clean bridge but remains below the current clean best.",
     "K4 train-state residual retrieval, scale 0.40, safe residuals, mean-by-type tangent consensus improves the clean bridge but is not yet the main result.",
-    "Policy-baseline abstention inside the same-state no-expert lattice drops from 56.99% to 40.70%, so the mechanism result should emphasize counterfactual proposal geometry rather than policy fallback."
+    "Policy-baseline abstention inside the same-state no-expert lattice drops from 56.99% to 40.70%, so the mechanism result should emphasize counterfactual proposal geometry rather than policy fallback.",
+    "Singleton near-miss priors 0.01/0.02 tie the clean best rather than improving it; the compatible chart preserves high-precision near-miss tangents, but they remain sparse."
   ]
 }

results/paper_table_status.md

@@ -71,8 +71,10 @@ Baseline h=16 policy: 29.74%
 | 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_residual_k4_composemasked_grid035040045 | K4 composed type-consensus residual retrieval, masked, scales 0.35/0.40/0.45, margin 0.20 | complete | 35.30% | +5.57 pp | yes | no | no | local tangent composition with anti-goal composite masks |
 | retrieval_residual_k4_composemasked_grid035040045_noopbonus003 | K4 composed type-consensus residual retrieval, masked, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03 | complete | 35.54% | +5.80 pp | yes | no | no | local tangent composition with anti-goal composite masks on the current best typed prior |
-| retrieval_residual_k4_composemasked_dropnmnoop_grid035040045 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite | pending 14927919/14927921 | pending | pending | yes | no | no | exact compatibility mask without typed no-op prior |
+| retrieval_residual_k4_composemasked_dropnmnoop_grid035040045 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite | complete | 35.48% | +5.74 pp | yes | no | no | exact compatibility mask without typed no-op prior |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite | complete | 35.59% | +5.86 pp | yes | no | no | exact incompatibility mask for one weak composed tangent pair |
+| retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001 | K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01 | complete | 35.59% | +5.86 pp | yes | no | no | revive high-precision singleton near-miss tangents without boosting toxic near-miss+no-op composites |
+| retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002 | K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.02 | complete | 35.59% | +5.86 pp | yes | no | no | stronger singleton near-miss revival prior on the compatible local tangent chart |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, source-score bonus 0.02 | complete | 35.48% | +5.74 pp | yes | no | no | train-measured source-score prior on the compatible local tangent chart |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_srcscorebonus002 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, no-op bonus 0.03, source-score bonus 0.02 | complete | 35.54% | +5.80 pp | yes | no | no | train-measured source-score calibration on the current compatible typed-prior row |
 | retrieval_residual_k4_composemasked_dropnmnoop_l2comp002_grid035040045_noopbonus003 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, composite L2 penalty 0.02 | complete | 35.54% | +5.80 pp | yes | no | no | compatible local tangent composition with trust-radius regularization |
@@ -146,3 +148,4 @@ Baseline h=16 policy: 29.74%
 - KNN counterfactual residual retrieval is a positive clean bridge but remains below the current clean best.
 - K4 train-state residual retrieval, scale 0.40, safe residuals, mean-by-type tangent consensus improves the clean bridge but is not yet the main result.
 - Policy-baseline abstention inside the same-state no-expert lattice drops from 56.99% to 40.70%, so the mechanism result should emphasize counterfactual proposal geometry rather than policy fallback.
+- Singleton near-miss priors 0.01/0.02 tie the clean best rather than improving it; the compatible chart preserves high-precision near-miss tangents, but they remain sparse.

scripts/build_paper_analysis.py

@@ -369,6 +369,22 @@ METHODS = [
             "k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_summary.json"
         ),
     ),
+    MethodSpec(
+        key="residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
+        label="K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.01",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_"
+            "k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p01_summary.json"
+        ),
+    ),
+    MethodSpec(
+        key="residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002",
+        label="K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.02",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_"
+            "k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p02_summary.json"
+        ),
+    ),
     MethodSpec(
         key="residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002",
         label="K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, source-score bonus 0.02",

scripts/build_paper_table_status.py

@@ -715,6 +715,26 @@ SPECS = [
         story_role="exact incompatibility mask for one weak composed tangent pair",
         pending_job="14915009/14915013",
     ),
+    ResultSpec(
+        key="retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
+        label="K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01",
+        path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p01_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="revive high-precision singleton near-miss tangents without boosting toxic near-miss+no-op composites",
+        pending_job="14929087/14929152",
+    ),
+    ResultSpec(
+        key="retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002",
+        label="K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.02",
+        path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_safe_margin0p20_noopbonus0p03_nmbonus0p02_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="stronger singleton near-miss revival prior on the compatible local tangent chart",
+        pending_job="14929120/14929175",
+    ),
     ResultSpec(
         key="retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002",
         label="K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, source-score bonus 0.02",
@@ -1460,6 +1480,16 @@ def _decision_notes(rows: list[dict[str, Any]]) -> list[str]:
             f"{_fmt_percent(no_expert['success'])} to {_fmt_percent(policy_baseline['success'])}, "
             "so the mechanism result should emphasize counterfactual proposal geometry rather than policy fallback."
         )
+    nm_ties = [
+        by_key["retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001"],
+        by_key["retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002"],
+    ]
+    if clean_best is not None and all(row["success"] is not None for row in nm_ties):
+        if all(abs(row["success"] - clean_best["success"]) < 1.0e-12 for row in nm_ties):
+            notes.append(
+                "Singleton near-miss priors 0.01/0.02 tie the clean best rather than improving it; "
+                "the compatible chart preserves high-precision near-miss tangents, but they remain sparse."
+            )
     return notes
 
 

scripts/eval_maniskill_policy_rollout.py

@@ -252,6 +252,14 @@ def main(argv: list[str] | None = None) -> int:
         help="Let composite candidate types inherit the sum of configured component bonuses "
         "unless an exact composite bonus is configured.",
     )
+    parser.add_argument(
+        "--candidate-oracle-rollouts",
+        type=int,
+        default=0,
+        help="Diagnostic only: execute the selected candidate plus the top field-ranked "
+        "residual candidates and report the best measured outcome in that prefix. "
+        "Zero disables the diagnostic and preserves deployment evaluation.",
+    )
     args = parser.parse_args(argv)
     lattice_exclude_types = tuple(
         item.strip() for item in args.lattice_exclude_types.split(",") if item.strip()
@@ -326,6 +334,7 @@ def main(argv: list[str] | None = None) -> int:
         lattice_exclude_types=lattice_exclude_types,
         candidate_type_bonuses=candidate_type_bonuses,
         candidate_type_bonus_components=args.candidate_type_bonus_components,
+        candidate_oracle_rollouts=args.candidate_oracle_rollouts,
     )
     print(json.dumps({key: value for key, value in result.items() if key != "rows"}, indent=2))
     return 0

scripts/slurm/eval_maniskill_policy_rollout.sbatch

@@ -79,6 +79,7 @@ if [[ -n "${CANDIDATE_TYPE_BONUSES_COLON:-}" ]]; then
   CANDIDATE_TYPE_BONUSES="${CANDIDATE_TYPE_BONUSES_COLON//:/,}"
 fi
 CANDIDATE_TYPE_BONUS_COMPONENTS="${CANDIDATE_TYPE_BONUS_COMPONENTS:-0}"
+CANDIDATE_ORACLE_ROLLOUTS="${CANDIDATE_ORACLE_ROLLOUTS:-0}"
 
 module load StdEnv/2023 apptainer/1.4.5
 cd "$PROJECT_DIR"
@@ -148,4 +149,5 @@ apptainer exec --nv \
     --retrieval-residual-reduce "$RETRIEVAL_RESIDUAL_REDUCE" \
     --lattice-exclude-types "$LATTICE_EXCLUDE_TYPES" \
     --candidate-type-bonuses "$CANDIDATE_TYPE_BONUSES" \
+    --candidate-oracle-rollouts "$CANDIDATE_ORACLE_ROLLOUTS" \
     "${EXTRA_ARGS[@]}"

tests/test_maniskill_policy_rollout.py

@@ -18,6 +18,7 @@ from dovla_cil.eval.maniskill_policy_rollout import (
     _RolloutCase,
     _adapt_action_dim,
     _attach_retrieved_residual_candidates,
+    _diagnostic_candidate_indices,
     _effective_lattice_candidate_count,
     _lattice_candidate_mask,
     _lattice_candidate_type_bonus,
@@ -79,6 +80,63 @@ def test_policy_rollout_summary_uses_measured_rollout_rows() -> None:
     assert summary["restore_max_error"] == 2e-7
 
 
+def test_policy_rollout_summary_includes_candidate_oracle_when_present() -> None:
+    rows = [
+        {
+            "success": False,
+            "progress": 0.1,
+            "oracle_success": True,
+            "expert_success": True,
+            "oracle_regret": 1.9,
+            "expert_regret": 1.9,
+            "action_mse_to_best": 0.5,
+            "restore_error": 1e-7,
+            "candidate_oracle_success": True,
+            "candidate_oracle_progress": 0.8,
+            "candidate_oracle_score": 1.8,
+            "candidate_oracle_regret": 0.2,
+            "candidate_oracle_score_gain_over_selected": 1.7,
+            "candidate_oracle_improves_selected": True,
+            "candidate_oracle_selected_branch_success": False,
+            "candidate_oracle_selected_branch_progress": 0.1,
+            "candidate_oracle_restore_error": 3e-7,
+            "candidate_oracle_candidate_type": "retrieval_residual_residual_no_op",
+        },
+        {
+            "success": True,
+            "progress": 0.9,
+            "oracle_success": True,
+            "expert_success": True,
+            "oracle_regret": 0.0,
+            "expert_regret": 0.0,
+            "action_mse_to_best": 0.0,
+            "restore_error": 2e-7,
+            "candidate_oracle_success": True,
+            "candidate_oracle_progress": 0.9,
+            "candidate_oracle_score": 1.9,
+            "candidate_oracle_regret": 0.0,
+            "candidate_oracle_score_gain_over_selected": 0.0,
+            "candidate_oracle_improves_selected": False,
+            "candidate_oracle_selected_branch_success": True,
+            "candidate_oracle_selected_branch_progress": 0.9,
+            "candidate_oracle_restore_error": 2e-7,
+            "candidate_oracle_candidate_type": "retrieval_residual_policy_residual",
+        },
+    ]
+
+    summary = _summarize_rows(rows)
+
+    assert summary["candidate_oracle_success_rate"] == 1.0
+    assert np.isclose(summary["candidate_oracle_progress"], 0.85)
+    assert np.isclose(summary["candidate_oracle_score_gain_over_selected"], 0.85)
+    assert summary["candidate_oracle_improvement_rate"] == 0.5
+    assert summary["candidate_oracle_restore_max_error"] == 3e-7
+    assert summary["candidate_oracle_type_counts"] == {
+        "retrieval_residual_residual_no_op": 1,
+        "retrieval_residual_policy_residual": 1,
+    }
+
+
 def test_policy_rollout_loads_state_archive(tmp_path: Path) -> None:
     archive = {"format": "dovla_maniskill_state_archive", "initial": {"g0": {"actors": {}}}}
     (tmp_path / "state_archive.pkl").write_bytes(pickle.dumps(archive))
@@ -230,6 +288,27 @@ def test_retrieval_residual_margin_can_abstain_to_policy() -> None:
     assert index.tolist() == [0]
 
 
+def test_candidate_oracle_prefix_keeps_deployed_candidate_first() -> None:
+    import torch
+
+    potentials = torch.tensor(
+        [
+            [0.1, 0.3, 0.2],
+            [float("-inf"), 0.4, 0.5],
+        ],
+        dtype=torch.float32,
+    )
+
+    indices = _diagnostic_candidate_indices(
+        potentials,
+        np.asarray([0, 1], dtype=np.int64),
+        candidate_oracle_rollouts=3,
+        torch=torch,
+    )
+
+    assert indices.tolist() == [[0, 1, 2], [1, 2, 1]]
+
+
 def test_field_mode_scores_clamped_candidates() -> None:
     import torch