Auto-sync: 2026-06-28 23:54:44 (part 2)

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_consensus_noopbonus003",
-  "generated_utc": "2026-06-29T03:45:52+00:00",
+  "generated_utc": "2026-06-29T03:58:54+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.06956521739130428,
     "best_clean_vs_h16": 0.05507246376811592,
@@ -1351,14 +1351,158 @@
       "std_success": 0.012049049096131323
     },
     "residual_k4_consensus_srcprogbonus003": {
+      "ci95_success": 0.03518915067855206,
       "label": "K4 mean-by-type tangent consensus, source-progress bonus 0.03",
-      "missing": true,
-      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p03_summary.json"
+      "mean_action_mse_to_best": 0.39541252239431807,
+      "mean_progress": 0.5668036828658902,
+      "mean_success": 0.352463768115942,
+      "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": 0.3244027510331859,
+          "std_success": 0.026822320875242476
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20103794840636946,
+          "std_success": 0.014667153114049147
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7478514959028968,
+          "std_success": 0.07707721123855243
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.2011137011137011,
+          "std_success": 0.07105663963231741
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3816248880099991,
+        "1": 0.38801324410445015,
+        "2": 0.41659943506850494
+      },
+      "seed_progress": {
+        "0": 0.5509942502163998,
+        "1": 0.5665316749817652,
+        "2": 0.5828851233995058
+      },
+      "seed_success": {
+        "0": 0.34608695652173915,
+        "1": 0.3426086956521739,
+        "2": 0.36869565217391304
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1652,
+        "retrieval_residual_residual_no_op": 46,
+        "retrieval_residual_residual_wrong_gripper": 27
+      },
+      "selected_residual_scale_counts": {
+        "0.4": 1725
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1652.0,
+          "mean_progress": 0.5610079844250168,
+          "success_rate": 0.3450363196125908
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 46.0,
+          "mean_progress": 0.7675236147706923,
+          "success_rate": 0.5652173913043478
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 27.0,
+          "mean_progress": 0.5794472738548562,
+          "success_rate": 0.4444444444444444
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p03_summary.json",
+      "std_success": 0.0141643962004297
     },
     "residual_k4_consensus_srcprogbonus005": {
+      "ci95_success": 0.03374487249808164,
       "label": "K4 mean-by-type tangent consensus, source-progress bonus 0.05",
-      "missing": true,
-      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p05_summary.json"
+      "mean_action_mse_to_best": 0.3956964282710375,
+      "mean_progress": 0.5664844556027748,
+      "mean_success": 0.35130434782608694,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.2599364286510761,
+          "std_success": 0.054567202237336666
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.3228001869306218,
+          "std_success": 0.025410290531060663
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20103794840636946,
+          "std_success": 0.014667153114049147
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7478514959028968,
+          "std_success": 0.07707721123855243
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.2011137011137011,
+          "std_success": 0.07105663963231741
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3820005126459443,
+        "1": 0.38828682124776687,
+        "2": 0.41680195091940136
+      },
+      "seed_progress": {
+        "0": 0.5506076139916221,
+        "1": 0.5665060519297486,
+        "2": 0.5823397008869537
+      },
+      "seed_success": {
+        "0": 0.3443478260869565,
+        "1": 0.3426086956521739,
+        "2": 0.36695652173913046
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1639,
+        "retrieval_residual_residual_no_op": 55,
+        "retrieval_residual_residual_wrong_gripper": 31
+      },
+      "selected_residual_scale_counts": {
+        "0.4": 1725
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1639.0,
+          "mean_progress": 0.5596922661425185,
+          "success_rate": 0.34350213544844416
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 55.0,
+          "mean_progress": 0.7515055749904026,
+          "success_rate": 0.5272727272727272
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 31.0,
+          "mean_progress": 0.597330809120209,
+          "success_rate": 0.45161290322580644
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p05_summary.json",
+      "std_success": 0.01358304291462029
     },
     "residual_k4_consensus_wgbonus003": {
       "ci95_success": 0.03271496576240179,

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-29T03:45:52+00:00`
+Generated: `2026-06-29T03:58:54+00:00`
 
 ## Main Seed Statistics
 
@@ -26,8 +26,8 @@ Generated: `2026-06-29T03:45:52+00:00`
 | residual_k4_consensus_noopbonus003_srcprog025 | K4 mean-by-type tangent consensus, no-op bonus 0.03, source progress >= 0.25 | 3 | 35.19% +/- 1.32 | +/- 3.27 | 56.69% | 0.396 | +5.45 pp |
 | residual_k4_consensus_noopbonus003_srcprog050 | K4 mean-by-type tangent consensus, no-op bonus 0.03, source progress >= 0.50 | 3 | 34.96% +/- 1.55 | +/- 3.84 | 56.53% | 0.396 | +5.22 pp |
 | residual_k4_consensus_noopbonus003_srcprog075 | K4 mean-by-type tangent consensus, no-op bonus 0.03, source progress >= 0.75 | 3 | 34.72% +/- 1.42 | +/- 3.52 | 56.44% | 0.396 | +4.99 pp |
-| residual_k4_consensus_srcprogbonus003 | K4 mean-by-type tangent consensus, source-progress bonus 0.03 | 0 | missing | missing | missing | missing | missing |
-| residual_k4_consensus_srcprogbonus005 | K4 mean-by-type tangent consensus, source-progress bonus 0.05 | 0 | missing | missing | missing | missing | missing |
+| residual_k4_consensus_srcprogbonus003 | K4 mean-by-type tangent consensus, source-progress bonus 0.03 | 3 | 35.25% +/- 1.42 | +/- 3.52 | 56.68% | 0.395 | +5.51 pp |
+| residual_k4_consensus_srcprogbonus005 | K4 mean-by-type tangent consensus, source-progress bonus 0.05 | 3 | 35.13% +/- 1.36 | +/- 3.37 | 56.65% | 0.396 | +5.39 pp |
 | residual_taskrelative_k4_consensus_noopbonus003 | K4 task-relative tangent consensus, no-op bonus 0.03 | 3 | 34.43% +/- 1.22 | +/- 3.02 | 56.19% | 0.399 | +4.70 pp |
 | residual_k4_consensus_noopbonus001 | K4 mean-by-type tangent consensus, no-op bonus 0.01 | 3 | 35.19% +/- 1.32 | +/- 3.27 | 56.63% | 0.395 | +5.45 pp |
 | residual_k4_consensus_noopbonus002 | K4 mean-by-type tangent consensus, no-op bonus 0.02 | 3 | 35.19% +/- 1.32 | +/- 3.27 | 56.64% | 0.395 | +5.45 pp |

results/paper_core_results.md

@@ -41,6 +41,8 @@ and the remaining clean-to-same-state proposal gap is `+21.74 pp`.
 | K2 task-relative residual retrieval, safe residuals + advantage margin 0.20 | No | No | 34.26% | +4.52 pp | Actor-pose-only retrieval is too lossy; raw full-state similarity is better for residual transfer |
 | K4 train-state residual retrieval, safe residuals + mean-by-type tangent consensus | No | No | 34.96% | +5.22 pp | Near-tie clean diagnostic; consensus alone does not beat raw K2 residuals |
 | K4 mean-by-type residual retrieval + no-op prior 0.03 | No | No | 35.25% | +5.51 pp | Current best clean diagnostic; 0.025-0.035 forms a small plateau that nudges high-value no-op residuals without changing the core proposal family |
+| K4 mean-by-type residual retrieval + source-progress prior 0.03 | No | No | 35.25% | +5.51 pp | Ties the current best exactly without a hand typed no-op prior; train-measured source progress can replace but not improve the typed prior |
+| K4 mean-by-type residual retrieval + source-progress prior 0.05 | No | No | 35.13% | +5.39 pp | A stronger measured-progress prior over-selects nonzero residuals and drops below the plateau |
 | K4 mean-by-type residual retrieval + no-op prior + source-progress gate | No | No | 34.72-35.19% | +4.99-5.45 pp | Train-source progress viability is a near-tie/negative gate; soft threshold 0.25 is closest, while stricter thresholds over-abstain below the no-op plateau |
 | K4 task-relative mean-by-type residual retrieval + no-op prior 0.03 | No | No | 34.43% | +4.70 pp | Task-relative target/reference pose retrieval underperforms the raw-metric no-op plateau |
 | K4 kernel-weighted residual consensus + no-op prior 0.03 | No | No | 35.13-35.19% | +5.39-5.45 pp | Distance-weighted tangent interpolation is plausible but does not beat equal mean-consensus no-op plateau |
@@ -82,28 +84,30 @@ Suggested main-table rows:
 11. K2 train-state residual retrieval, typed safe families + advantage margin 0.20
 12. K4 train-state residual retrieval, mean-by-type tangent consensus
 13. K4 mean-by-type residual retrieval + no-op prior plateau, canonical 0.03
-14. Source-progress viability gate diagnostics
-15. K2/K4 task-relative retrieval metric diagnostics
-16. K4 kernel-weighted residual consensus + no-op prior diagnostics
-17. K4 field-softmax residual barycenter + margin diagnostics
-18. K4 mean-by-type residual retrieval + wrong-gripper typed-prior diagnostics
-19. K2 broad tangent ray-search
-20. Residual-tangent distillation policy
-21. Residual+Gaussian hybrid, K32 sigma0.35
-22. Lattice, near-miss only
-23. Lattice, no expert
-24. Lattice, no expert + policy baseline candidate
-25. Lattice, full
-26. Oracle ceiling
+14. K4 mean-by-type residual retrieval + source-progress prior diagnostics
+15. Source-progress viability gate diagnostics
+16. K2/K4 task-relative retrieval metric diagnostics
+17. K4 kernel-weighted residual consensus + no-op prior diagnostics
+18. K4 field-softmax residual barycenter + margin diagnostics
+19. K4 mean-by-type residual retrieval + wrong-gripper typed-prior diagnostics
+20. K2 broad tangent ray-search
+21. Residual-tangent distillation policy
+22. Residual+Gaussian hybrid, K32 sigma0.35
+23. Lattice, near-miss only
+24. Lattice, no expert
+25. Lattice, no expert + policy baseline candidate
+26. Lattice, full
+27. Oracle ceiling
 
 Suggested claim:
 
 > DoVLA-CIL is not a better behavior-cloning policy; it is a local counterfactual action
 > selection rule. Deployment-clean K4 consensus residual transport with advantage
-> abstention and a small typed no-op prior plateau gives the strongest clean gain so far, while ungated KNN residual
+> abstention and either a small typed no-op prior or an equivalent train-source
+> progress prior gives the strongest clean gain so far, while ungated KNN residual
 > retrieval, field-gradient ascent, broader non-expert BC targets, field-teacher/tangent distillation, z-score/task-relative retrieval,
 > train-family reliability priors, policy-relative anchoring, residual+Gaussian hybrids,
-> source-progress viability gates, tangent consensus, kernel-weighted tangent interpolation, field-softmax tangent barycenters, tangent ray-search, wrong-gripper typed priors, and same-state policy-baseline fallback fail to improve the main rows.
+> source-progress viability gates, overly strong progress priors, tangent consensus, kernel-weighted tangent interpolation, field-softmax tangent barycenters, tangent ray-search, wrong-gripper typed priors, and same-state policy-baseline fallback fail to improve the main rows.
 > The large effect appears only when the field is queried on
 > same-state intervention proposals, and the mechanism is isolated to local near-miss
 > counterfactual geometry.

results/paper_story_memo.md

@@ -38,6 +38,7 @@ when queried on proposal geometry that matches those local counterfactuals.
 | Z-score retrieval metric does not help | z-score rows reach 32.23-32.81%, below raw retrieval | Negative diagnostic |
 | Task-relative actor-pose retrieval metric does not improve tangent transfer | K2 task-relative residual retrieval reaches 34.26% vs raw K2 35.01%; K4 task-relative mean-by-type + no-op reaches 34.43% vs raw K4 35.25% | Negative diagnostic |
 | Train-source progress viability is too blunt a residual gate | source-progress thresholds 0.25/0.50/0.75 reach 35.19%/34.96%/34.72%, below the unfiltered no-op plateau at 35.25% | Negative/near-tie diagnostic |
+| Continuous train-source progress prior can replace the typed no-op prior but not improve it | source-progress bonus 0.03 ties the 35.25% best exactly; bonus 0.05 drops to 35.13% | Cleaner tie diagnostic |
 | Train-split residual family reliability does not recover the typed mask | after fixing threshold pass-through, scale-0.35 thresholds 0.10/0.25 reach 33.33%/33.28%, below typed safe residuals | Negative diagnostic |
 | Residual-tangent distillation does not solve clean proposal generation | aligned allmap tangent student reaches 28.87% despite low pseudo-target BC loss | Negative diagnostic |
 | Policy-relative residual anchoring does not improve the bridge | policy-anchor safe residual transport ties 33.74% rather than improving expert-anchor residuals | Negative diagnostic |
@@ -63,23 +64,24 @@ clean proposal result, the intended main rows are:
 13. K2 train-state residual retrieval, typed safe families + advantage margin: 35.01%
 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. Source-progress viability gates: 35.19% / 34.96% / 34.72% for thresholds 0.25 / 0.50 / 0.75
-17. K4 kernel-weighted tangent consensus / + no-op prior: 34.96% / 35.19%
-18. K4 field-softmax tangent transport / best margin sweep: 34.96% / 35.19%
-19. Wrong-gripper prior / no-op+wrong-gripper prior: 35.19% / 35.25%
-20. K2 broad tangent ray-search: 34.96%
-21. K1/K2 tight tangent ray-search: 34.84% / 34.84%
-22. K4 tight tangent ray-search: 34.55%
-23. Residual-tangent distillation policy: 28.87%
-24. Z-score residual retrieval: 32.23-32.81%
-25. Task-relative residual retrieval metric: 34.26-34.43%
-26. Train-family reliability prior: 33.28-33.33%
-27. Residual+Gaussian hybrid K32/K64: 31.30% / 30.90%
-28. Lattice, near-miss only: 55.94%
-29. Lattice, no expert: 56.99%
-30. Lattice, no expert + policy baseline candidate: 40.70%
-31. Lattice, full: 69.33%
-32. Oracle ceiling: 86.78%
+16. K4 mean-by-type tangent consensus + train-source progress prior: 35.25% at bonus 0.03; 35.13% at bonus 0.05
+17. Source-progress viability gates: 35.19% / 34.96% / 34.72% for thresholds 0.25 / 0.50 / 0.75
+18. K4 kernel-weighted tangent consensus / + no-op prior: 34.96% / 35.19%
+19. K4 field-softmax tangent transport / best margin sweep: 34.96% / 35.19%
+20. Wrong-gripper prior / no-op+wrong-gripper prior: 35.19% / 35.25%
+21. K2 broad tangent ray-search: 34.96%
+22. K1/K2 tight tangent ray-search: 34.84% / 34.84%
+23. K4 tight tangent ray-search: 34.55%
+24. Residual-tangent distillation policy: 28.87%
+25. Z-score residual retrieval: 32.23-32.81%
+26. Task-relative residual retrieval metric: 34.26-34.43%
+27. Train-family reliability prior: 33.28-33.33%
+28. Residual+Gaussian hybrid K32/K64: 31.30% / 30.90%
+29. Lattice, near-miss only: 55.94%
+30. Lattice, no expert: 56.99%
+31. Lattice, no expert + policy baseline candidate: 40.70%
+32. Lattice, full: 69.33%
+33. Oracle ceiling: 86.78%
 
 ## Novelty Framing
 
@@ -107,9 +109,9 @@ test-time search. The cleaner novelty is:
 
 ## Job Status
 
-Last checked: `2026-06-29 03:38 UTC`. The source-progress viability-gate batch
+Last checked: `2026-06-29 03:58 UTC`. The source-progress bonus-prior batch
 completed after passing CPU/unit smokes, and the paper table/paired analysis now
-include thresholds 0.25, 0.50, and 0.75.
+include both threshold gates and continuous train-source progress priors.
 
 - `14858328`-`14858333`: completed train-split `field_selected_noexpert_bc5`;
   direct rollout is 26.84%, field-guided best is 27.65%.
@@ -235,6 +237,13 @@ include thresholds 0.25, 0.50, and 0.75.
   after the first `0.50` seeds appeared overly conservative. It reaches 35.19%,
   still below the 35.25% no-op plateau; summary `14894439` and rebuild
   `14894440` completed.
+- `14894672`/`14894673`: completed unit and CPU rollout smokes for the continuous
+  train-source progress bonus path. The unit smoke validated bonus padding
+  (`source_progress_bonuses == [[0, 0.08, 0], [0, 0.08, 0.08]]`).
+- `14894674`/`14894675`: completed source-progress bonus arrays with no fixed
+  no-op prior. Bonus `0.03` ties the current best at 35.25%; bonus `0.05`
+  reaches 35.13%. Summary jobs `14894676`/`14894677` completed; rebuild job
+  `14894678` was queued after them.
 - `14869627`: completed CPU Apptainer smoke for the new residual scale-grid
   selector. It selected index `3` on a two-residual/two-scale toy case and
   returned the expected action `0.20`, validating the candidate expansion and
@@ -257,8 +266,10 @@ include thresholds 0.25, 0.50, and 0.75.
 - Use K4 mean-by-type residual transport with advantage abstention and a small
   typed no-op prior plateau (0.025-0.035, canonical midpoint 0.03, 35.25%) as
   the current best clean deployment diagnostic, not as a SOTA claim. The
-  completed K2/ray-search rows are near-ties that support the sparse-intervention
-  story.
+  continuous train-source progress prior at bonus 0.03 ties this result without
+  a hand typed no-op prior, which is a cleaner story hook but still not a higher
+  SOTA number. The completed K2/ray-search rows are near-ties that support the
+  sparse-intervention story.
 - Use `results/paper_analysis.md` for paired seed deltas, per-task gaps, and
   selection histograms when writing reviewer-facing tables.
 - Treat z-score and task-relative retrieval metrics, source-progress viability gates,

results/paper_table_status.json

@@ -907,14 +907,14 @@
       "story_role": "train-source progress prior replacing typed no-op prior",
       "fallback_success": null,
       "pending_job": "14894674/14894676",
-      "path_exists": false,
-      "status": "pending",
-      "success": null,
-      "std_success": null,
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35246376811594204,
+      "std_success": 0.014164396200429698,
       "completed_seeds": null,
-      "num_completed": null,
+      "num_completed": 3,
       "best_config": null,
-      "gain_vs_h16_policy": null
+      "gain_vs_h16_policy": 0.055072463768115976
     },
     {
       "key": "retrieval_residual_k4_mean_srcprogbonus005",
@@ -926,14 +926,14 @@
       "story_role": "strong train-source progress prior replacing typed no-op prior",
       "fallback_success": null,
       "pending_job": "14894675/14894677",
-      "path_exists": false,
-      "status": "pending",
-      "success": null,
-      "std_success": null,
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35130434782608694,
+      "std_success": 0.013583042914620291,
       "completed_seeds": null,
-      "num_completed": null,
+      "num_completed": 3,
       "best_config": null,
-      "gain_vs_h16_policy": null
+      "gain_vs_h16_policy": 0.05391304347826087
     },
     {
       "key": "retrieval_residual_taskrelative_k4_mean_noopbonus003",

results/paper_table_status.md

@@ -50,8 +50,8 @@ Baseline h=16 policy: 29.74%
 | retrieval_residual_k4_mean_noopbonus003_srcprog025 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op bonus 0.03, source progress >= 0.25 | complete | 35.19% | +5.45 pp | yes | no | no | soft train-source viability gate for sparse residual transport |
 | retrieval_residual_k4_mean_noopbonus003_srcprog050 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op bonus 0.03, source progress >= 0.50 | complete | 34.96% | +5.22 pp | yes | no | no | train-source viability gate for sparse residual transport |
 | retrieval_residual_k4_mean_noopbonus003_srcprog075 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op bonus 0.03, source progress >= 0.75 | complete | 34.72% | +4.99 pp | yes | no | no | strict train-source viability gate for sparse residual transport |
-| retrieval_residual_k4_mean_srcprogbonus003 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, source-progress bonus 0.03 | pending 14894674/14894676 | pending | pending | yes | no | no | train-source progress prior replacing typed no-op prior |
-| retrieval_residual_k4_mean_srcprogbonus005 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, source-progress bonus 0.05 | pending 14894675/14894677 | pending | pending | yes | no | no | strong train-source progress prior replacing typed no-op prior |
+| retrieval_residual_k4_mean_srcprogbonus003 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, source-progress bonus 0.03 | complete | 35.25% | +5.51 pp | yes | no | no | train-source progress prior replacing typed no-op prior |
+| retrieval_residual_k4_mean_srcprogbonus005 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, source-progress bonus 0.05 | complete | 35.13% | +5.39 pp | yes | no | no | strong train-source progress prior replacing typed no-op prior |
 | retrieval_residual_taskrelative_k4_mean_noopbonus003 | K4 task-relative mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op residual bonus 0.03 | complete | 34.43% | +4.70 pp | yes | no | no | task-relative state metric for typed sparse-intervention prior |
 | retrieval_residual_k4_mean_noopbonus001 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op residual bonus 0.01 | complete | 35.19% | +5.45 pp | yes | no | no | typed sparse-intervention prior fine sweep |
 | retrieval_residual_k4_mean_noopbonus002 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op residual bonus 0.02 | complete | 35.19% | +5.45 pp | yes | no | no | typed sparse-intervention prior fine sweep |