Auto-sync: 2026-06-29 02:57:43 (part 3)

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_consensus_grid035040045_noopbonus003",
-  "generated_utc": "2026-06-29T06:47:32+00:00",
+  "generated_utc": "2026-06-29T07:06:31+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.07130434782608691,
     "best_clean_vs_h16": 0.05681159420289855,
@@ -507,6 +507,243 @@
       "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid035040045_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json",
       "std_success": 0.012049049096131323
     },
+    "residual_k4_consensus_grid035045055_noopbonus003": {
+      "ci95_success": 0.0228624037279942,
+      "label": "K4 mean-by-type tangent consensus, scales 0.35/0.45/0.55, no-op bonus 0.03",
+      "mean_action_mse_to_best": 0.4017242217439132,
+      "mean_progress": 0.5679968697053118,
+      "mean_success": 0.35130434782608694,
+      "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": 0.32313730955035297,
+          "std_success": 0.016291886617014133
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.19226601858180806,
+          "std_success": 0.029812623771358626
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7538059441366793,
+          "std_success": 0.07965457487204111
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.19740999740999743,
+          "std_success": 0.07397934256534576
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3897169904621399,
+        "1": 0.3928398724855936,
+        "2": 0.4226158022840062
+      },
+      "seed_progress": {
+        "0": 0.5549358964865299,
+        "1": 0.5698409637744012,
+        "2": 0.5792137488550044
+      },
+      "seed_success": {
+        "0": 0.3443478260869565,
+        "1": 0.34782608695652173,
+        "2": 0.3617391304347826
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1610,
+        "retrieval_residual_residual_no_op": 82,
+        "retrieval_residual_residual_wrong_gripper": 33
+      },
+      "selected_residual_scale_counts": {
+        "0.35": 1627,
+        "0.45": 21,
+        "0.55": 77
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1610.0,
+          "mean_progress": 0.5596588065308825,
+          "success_rate": 0.3440993788819876
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 82.0,
+          "mean_progress": 0.678117303995461,
+          "success_rate": 0.4268292682926829
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 33.0,
+          "mean_progress": 0.7011606908883109,
+          "success_rate": 0.5151515151515151
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid035045055_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+      "std_success": 0.009202613255876844
+    },
+    "residual_k4_consensus_grid040045050_noopbonus003": {
+      "ci95_success": 0.025316336466430018,
+      "label": "K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, no-op bonus 0.03",
+      "mean_action_mse_to_best": 0.3983363078623687,
+      "mean_progress": 0.5691949463564601,
+      "mean_success": 0.35362318840579715,
+      "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": 0.3264233753364188,
+          "std_success": 0.019140239285976537
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20103794840636946,
+          "std_success": 0.014667153114049147
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7541032711666796,
+          "std_success": 0.07644139692615382
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.19740999740999743,
+          "std_success": 0.07397934256534576
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3837931162734394,
+        "1": 0.39108053201080667,
+        "2": 0.42013527530285977
+      },
+      "seed_progress": {
+        "0": 0.5553471852506956,
+        "1": 0.5711554712316265,
+        "2": 0.581082182587058
+      },
+      "seed_success": {
+        "0": 0.34608695652173915,
+        "1": 0.34956521739130436,
+        "2": 0.3652173913043478
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1626,
+        "retrieval_residual_residual_no_op": 67,
+        "retrieval_residual_residual_wrong_gripper": 32
+      },
+      "selected_residual_scale_counts": {
+        "0.4": 1641,
+        "0.45": 12,
+        "0.5": 72
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1626.0,
+          "mean_progress": 0.5610487258769904,
+          "success_rate": 0.34501845018450183
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 67.0,
+          "mean_progress": 0.7180690015318678,
+          "success_rate": 0.5074626865671642
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 32.0,
+          "mean_progress": 0.671419721446,
+          "success_rate": 0.46875
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+      "std_success": 0.010190374394925756
+    },
+    "residual_k4_consensus_grid040045050_srcscorebonus002": {
+      "ci95_success": 0.0262811119411741,
+      "label": "K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, source-score bonus 0.02",
+      "mean_action_mse_to_best": 0.39819325285518303,
+      "mean_progress": 0.5689121059149256,
+      "mean_success": 0.35304347826086957,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.2604587983659236,
+          "std_success": 0.06317251148949075
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.3264233753364188,
+          "std_success": 0.019140239285976537
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.20103794840636946,
+          "std_success": 0.014667153114049147
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7541032711666796,
+          "std_success": 0.07644139692615382
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.19740999740999743,
+          "std_success": 0.07397934256534576
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.3833637685717448,
+        "1": 0.39096053219521826,
+        "2": 0.4202554577985859
+      },
+      "seed_progress": {
+        "0": 0.5549538400178047,
+        "1": 0.5707065432065207,
+        "2": 0.5810759345204513
+      },
+      "seed_success": {
+        "0": 0.34608695652173915,
+        "1": 0.34782608695652173,
+        "2": 0.3652173913043478
+      },
+      "selected_candidate_type_counts": {
+        "retrieval_residual_policy_residual": 1630,
+        "retrieval_residual_residual_no_op": 60,
+        "retrieval_residual_residual_wrong_gripper": 35
+      },
+      "selected_residual_scale_counts": {
+        "0.4": 1644,
+        "0.45": 12,
+        "0.5": 69
+      },
+      "selected_type_outcomes": {
+        "retrieval_residual_policy_residual": {
+          "count": 1630.0,
+          "mean_progress": 0.5593074301753868,
+          "success_rate": 0.34294478527607364
+        },
+        "retrieval_residual_residual_no_op": {
+          "count": 60.0,
+          "mean_progress": 0.7584823973476886,
+          "success_rate": 0.55
+        },
+        "retrieval_residual_residual_wrong_gripper": {
+          "count": 35.0,
+          "mean_progress": 0.6912379336144243,
+          "success_rate": 0.4857142857142857
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json",
+      "std_success": 0.0105787174439969
+    },
     "residual_k4_consensus_margin015_noopbonus003": {
       "ci95_success": 0.027439420289855042,
       "label": "K4 mean-by-type tangent consensus, margin 0.15, no-op bonus 0.03",

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-29T06:47:32+00:00`
+Generated: `2026-06-29T07:06:31+00:00`
 
 ## Main Seed Statistics
 
@@ -30,6 +30,9 @@ Generated: `2026-06-29T06:47:32+00:00`
 | residual_k4_consensus_margin025_srcscorebonus002 | K4 mean-by-type tangent consensus, margin 0.25, source-score bonus 0.02 | 3 | 34.84% +/- 1.41 | +/- 3.49 | 56.41% | 0.395 | +5.10 pp |
 | residual_k4_consensus_grid035040045_noopbonus003 | K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03 | 3 | 35.42% +/- 1.12 | +/- 2.78 | 56.87% | 0.397 | +5.68 pp |
 | residual_k4_consensus_grid035040045_srcscorebonus002 | K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, source-score bonus 0.02 | 3 | 35.30% +/- 1.20 | +/- 2.99 | 56.76% | 0.397 | +5.57 pp |
+| residual_k4_consensus_grid040045050_noopbonus003 | K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, no-op bonus 0.03 | 3 | 35.36% +/- 1.02 | +/- 2.53 | 56.92% | 0.398 | +5.62 pp |
+| residual_k4_consensus_grid040045050_srcscorebonus002 | K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, source-score bonus 0.02 | 3 | 35.30% +/- 1.06 | +/- 2.63 | 56.89% | 0.398 | +5.57 pp |
+| residual_k4_consensus_grid035045055_noopbonus003 | K4 mean-by-type tangent consensus, scales 0.35/0.45/0.55, no-op bonus 0.03 | 3 | 35.13% +/- 0.92 | +/- 2.29 | 56.80% | 0.402 | +5.39 pp |
 | residual_k4_consensus_nooponly_noopbonus003 | K4 mean-by-type tangent consensus, no-op-only residuals, no-op bonus 0.03 | 3 | 35.19% +/- 1.18 | +/- 2.94 | 56.57% | 0.394 | +5.45 pp |
 | residual_k4_consensus_nooponly_srcscorebonus002 | K4 mean-by-type tangent consensus, no-op-only residuals, source-score bonus 0.02 | 3 | 35.19% +/- 1.18 | +/- 2.94 | 56.55% | 0.394 | +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 |

results/paper_core_results.md

@@ -46,6 +46,8 @@ no-op prior is `+5.68 pp` over canonical h=16, same-state no-expert lattice is
 | 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 + source-score prior 0.015/0.020 | No | No | 35.25% | +5.51 pp | Full train reward score, including terminal success, also replaces the fixed-scale typed prior without improving it |
 | K4 mean-by-type residual retrieval + scale-grid source-score prior 0.02 | No | No | 35.30% | +5.57 pp | Measured train-source prior benefits from tangent length calibration but remains below the typed no-op scale-grid row |
+| K4 mean-by-type residual retrieval + upper scale-grid no-op prior 0.03 | No | No | 35.36% | +5.62 pp | Scales 0.40/0.45/0.50 nearly tie but do not beat the 0.35/0.40/0.45 row |
+| K4 mean-by-type residual retrieval + wide scale-grid no-op prior 0.03 | No | No | 35.13% | +5.39 pp | Including 0.55 over-extends the transported tangent and drops below the best local calibration |
 | K4 mean-by-type residual retrieval + source-score prior 0.025 | No | No | 35.19% | +5.45 pp | A stronger reward-score prior drops below the plateau |
 | K4 mean-by-type residual retrieval, no-op-only residuals | No | No | 35.19% | +5.45 pp | Removing wrong-gripper residuals loses one success versus the fixed-scale safe-family plateau; the core gain is sparse no-op/tangent repair, with wrong-gripper acting only as a marginal helper |
 | K4 mean-by-type residual retrieval + margin sweep around 0.20 | No | No | 34.84-35.25% | +5.10-5.51 pp | Margin 0.20 is a local abstention optimum for both typed no-op and source-score priors; 0.15 and 0.25 drop below the plateau |
@@ -91,29 +93,32 @@ Suggested main-table rows:
 12. K4 train-state residual retrieval, mean-by-type tangent consensus
 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 mean-by-type residual retrieval + source-progress/source-score prior diagnostics
-16. K4 mean-by-type residual retrieval + no-op-only family diagnostic
-17. K4 mean-by-type residual retrieval + abstention margin fine sweep
-18. Source-progress viability gate diagnostics
-19. K2/K4 task-relative retrieval metric diagnostics
-20. K4 kernel-weighted residual consensus + no-op prior diagnostics
-21. K4 field-softmax residual barycenter + margin diagnostics
-22. K4 mean-by-type residual retrieval + wrong-gripper typed-prior diagnostics
-23. K2 broad tangent ray-search
-24. Residual-tangent distillation policy
-25. Residual+Gaussian hybrid, K32 sigma0.35
-26. Lattice, near-miss only
-27. Lattice, no expert
-28. Lattice, no expert + policy baseline candidate
-29. Lattice, full
-30. Oracle ceiling
+15. K4 mean-by-type residual retrieval + upper/wide tangent-length diagnostics
+16. K4 mean-by-type residual retrieval + source-progress/source-score prior diagnostics
+17. K4 mean-by-type residual retrieval + no-op-only family diagnostic
+18. K4 mean-by-type residual retrieval + abstention margin fine sweep
+19. Source-progress viability gate diagnostics
+20. K2/K4 task-relative retrieval metric diagnostics
+21. K4 kernel-weighted residual consensus + no-op prior diagnostics
+22. K4 field-softmax residual barycenter + margin diagnostics
+23. K4 mean-by-type residual retrieval + wrong-gripper typed-prior diagnostics
+24. K2 broad tangent ray-search
+25. Residual-tangent distillation policy
+26. Residual+Gaussian hybrid, K32 sigma0.35
+27. Lattice, near-miss only
+28. Lattice, no expert
+29. Lattice, no expert + policy baseline candidate
+30. Lattice, full
+31. 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, a small typed no-op prior, and field-gated tangent length
-> calibration gives the strongest clean gain so far; train-source progress/reward-score
+> calibration over a narrow local scale grid gives the strongest clean gain so far;
+> extending the scale grid upward is near-tie/negative, so the effect is local
+> calibration rather than larger steps. Train-source progress/reward-score
 > priors provide cleaner fixed-scale ties but not the top row. 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,

results/paper_story_memo.md

@@ -28,7 +28,7 @@ when queried on proposal geometry that matches those local counterfactuals.
 | Counterfactual advantage abstention improves clean transport | requiring field advantage over the zero-residual policy raises typed residual transport to 34.84%, and K2 retrieval reaches 35.01% | Supported as the previous clean best |
 | Clean residual transport behaves like sparse intervention | `paper_analysis.md` shows the best clean row abstains to zero-residual policy on 94.6% of states, while selected nonzero residuals succeed at 50.0% vs 34.58% 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% | Current best clean result |
+| 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% | Current best clean result; local scale calibration, not a larger-step effect |
 | Kernel-weighted tangent interpolation does not beat equal consensus | K4 kernel-weighted residual consensus reaches 34.96%; with no-op prior and scales 0.35/0.40/0.45 it reaches 35.13%/35.19%/35.19%, below the 35.25% mean-consensus plateau | Negative/near-tie diagnostic |
 | Field-conditioned tangent barycenters identify good sparse corrections but do not close the proposal gap | K4 field-softmax transport reaches 34.96%; with no-op prior and margins 0.10/0.05/0.00 it reaches 35.19%/35.07%/34.84%. Selected aggregate residuals are high-value (up to 60.00% success), but selecting more of them degrades the global row | Negative/near-tie diagnostic |
 | Tangent ray-search does not beat the typed-prior clean row | K1/K2 tight scale-grid ray search reach 34.84%; K2 broad reaches 34.96%; K4 tight reaches 34.55%, all below the scale-grid mean-consensus row at 35.42% | Near-tie/negative diagnostic |
@@ -69,27 +69,28 @@ 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 mean-by-type tangent consensus + train-source progress prior: 35.25% at bonus 0.03; 35.13% at bonus 0.05
-18. K4 mean-by-type tangent consensus + train-source reward-score prior: 35.25% at bonuses 0.015/0.020; 35.30% with scale grid; 35.19% at 0.025
-19. K4 mean-by-type tangent consensus, no-op-only residuals: 35.19% with either no-op bonus 0.03 or source-score bonus 0.02
-20. K4 mean-by-type abstention margin sweep: 35.07% / 35.25% / 34.84% for typed no-op margins 0.15 / 0.20 / 0.25; 34.96% / 35.25% / 34.84% for source-score margins
-21. Source-progress viability gates: 35.19% / 34.96% / 34.72% for thresholds 0.25 / 0.50 / 0.75
-22. K4 kernel-weighted tangent consensus / + no-op prior: 34.96% / 35.19%
-23. K4 field-softmax tangent transport / best margin sweep: 34.96% / 35.19%
-24. Wrong-gripper prior / no-op+wrong-gripper prior: 35.19% / 35.25%
-25. K2 broad tangent ray-search: 34.96%
-26. K1/K2 tight tangent ray-search: 34.84% / 34.84%
-27. K4 tight tangent ray-search: 34.55%
-28. Residual-tangent distillation policy: 28.87%
-29. Z-score residual retrieval: 32.23-32.81%
-30. Task-relative residual retrieval metric: 34.26-34.43%
-31. Train-family reliability prior: 33.28-33.33%
-32. Residual+Gaussian hybrid K32/K64: 31.30% / 30.90%
-33. Lattice, near-miss only: 55.94%
-34. Lattice, no expert: 56.99%
-35. Lattice, no expert + policy baseline candidate: 40.70%
-36. Lattice, full: 69.33%
-37. Oracle ceiling: 86.78%
+17. 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
+18. K4 mean-by-type tangent consensus + train-source progress prior: 35.25% at bonus 0.03; 35.13% at bonus 0.05
+19. K4 mean-by-type tangent consensus + train-source reward-score prior: 35.25% at bonuses 0.015/0.020; 35.30% with scale grid; 35.19% at 0.025
+20. K4 mean-by-type tangent consensus, no-op-only residuals: 35.19% with either no-op bonus 0.03 or source-score bonus 0.02
+21. K4 mean-by-type abstention margin sweep: 35.07% / 35.25% / 34.84% for typed no-op margins 0.15 / 0.20 / 0.25; 34.96% / 35.25% / 34.84% for source-score margins
+22. Source-progress viability gates: 35.19% / 34.96% / 34.72% for thresholds 0.25 / 0.50 / 0.75
+23. K4 kernel-weighted tangent consensus / + no-op prior: 34.96% / 35.19%
+24. K4 field-softmax tangent transport / best margin sweep: 34.96% / 35.19%
+25. Wrong-gripper prior / no-op+wrong-gripper prior: 35.19% / 35.25%
+26. K2 broad tangent ray-search: 34.96%
+27. K1/K2 tight tangent ray-search: 34.84% / 34.84%
+28. K4 tight tangent ray-search: 34.55%
+29. Residual-tangent distillation policy: 28.87%
+30. Z-score residual retrieval: 32.23-32.81%
+31. Task-relative residual retrieval metric: 34.26-34.43%
+32. Train-family reliability prior: 33.28-33.33%
+33. Residual+Gaussian hybrid K32/K64: 31.30% / 30.90%
+34. Lattice, near-miss only: 55.94%
+35. Lattice, no expert: 56.99%
+36. Lattice, no expert + policy baseline candidate: 40.70%
+37. Lattice, full: 69.33%
+38. Oracle ceiling: 86.78%
 
 ## Novelty Framing
 
@@ -117,9 +118,10 @@ test-time search. The cleaner novelty is:
 
 ## Job Status
 
-Last checked: `2026-06-29 06:47 UTC`. The K4 mean-by-type scale-grid sweep
-completed and produced a new clean best, 35.42%, while the paper table/paired
-analysis now use that row as `best_clean_key`.
+Last checked: `2026-06-29 07:04 UTC`. The K4 mean-by-type scale-grid sweep
+completed and produced a new clean best, 35.42%, while upper/wide follow-ups
+completed without improving it. The paper table/paired analysis use that row as
+`best_clean_key`.
 
 - `14858328`-`14858333`: completed train-split `field_selected_noexpert_bc5`;
   direct rollout is 26.84%, field-guided best is 27.65%.
@@ -276,6 +278,12 @@ analysis now use that row as `best_clean_key`.
   no-op prior row reaches a new clean best, 35.42%; the source-score prior row
   reaches 35.30%. Summary jobs `14897990`/`14897991` completed; rebuild job
   `14897992` was submitted, and local rebuilds updated the paper artifacts.
+- `14898107`/`14898108`/`14898109`: completed upper and wide K4 mean-by-type
+  scale-grid follow-ups. The no-op upper grid `0.40/0.45/0.50` reaches 35.36%,
+  the source-score upper grid reaches 35.30%, and the no-op wide grid
+  `0.35/0.45/0.55` reaches 35.13%. Summary jobs `14898110`/`14898111`/
+  `14898112` and rebuild job `14898113` completed; the best clean row remains
+  the `0.35/0.40/0.45` no-op grid at 35.42%.
 - `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
@@ -300,10 +308,12 @@ analysis now use that row as `best_clean_key`.
   `0.35/0.40/0.45` as the current best clean deployment diagnostic, 35.42%, not
   as a SOTA claim. The fixed-scale no-op plateau remains 35.25%; continuous
   train-source progress/reward-score priors tie that fixed-scale row, and
-  scale-grid source-score reaches 35.30% but not the new best. The no-op-only
-  family ablation reaches 35.19%, so wrong-gripper residuals are a marginal
-  helper rather than the core mechanism. The margin fine sweep confirms `0.20`
-  is a local abstention optimum for both typed and measured train-outcome
+  scale-grid source-score reaches 35.30% but not the new best. Upper and wide
+  scale grids reach 35.36% and 35.13%, so the scale-grid evidence supports a
+  local tangent-length calibration, not a monotone larger-step claim. The
+  no-op-only family ablation reaches 35.19%, so wrong-gripper residuals are a
+  marginal helper rather than the core mechanism. The margin fine sweep confirms
+  `0.20` is a local abstention optimum for both typed and measured train-outcome
   priors. 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

results/paper_table_status.json

@@ -973,6 +973,63 @@
       "best_config": null,
       "gain_vs_h16_policy": 0.0556521739130435
     },
+    {
+      "key": "retrieval_residual_k4_mean_grid040045050_noopbonus003",
+      "label": "K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, no-op bonus 0.03",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "upper tangent-length sweep for sparse mean-consensus residual transport",
+      "fallback_success": null,
+      "pending_job": "14898107/14898110",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.3536231884057971,
+      "std_success": 0.010190374394925756,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.05623188405797103
+    },
+    {
+      "key": "retrieval_residual_k4_mean_grid040045050_srcscorebonus002",
+      "label": "K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, source-score bonus 0.02",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "upper tangent-length sweep for measured-prior mean-consensus residual transport",
+      "fallback_success": null,
+      "pending_job": "14898108/14898111",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35304347826086957,
+      "std_success": 0.0105787174439969,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.0556521739130435
+    },
+    {
+      "key": "retrieval_residual_k4_mean_grid035045055_noopbonus003",
+      "label": "K4 mean-by-type residual retrieval, scales 0.35/0.45/0.55, margin 0.20, no-op bonus 0.03",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid035045055_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "wide tangent-length sweep for sparse mean-consensus residual transport",
+      "fallback_success": null,
+      "pending_job": "14898109/14898112",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.35130434782608694,
+      "std_success": 0.009202613255876844,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.05391304347826087
+    },
     {
       "key": "retrieval_residual_k4_mean_nooponly_noopbonus003",
       "label": "K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, no-op bonus 0.03",

results/paper_table_status.md

@@ -54,6 +54,9 @@ Baseline h=16 policy: 29.74%
 | retrieval_residual_k4_mean_margin025_srcscorebonus002 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.25, source-score bonus 0.02 | complete | 34.84% | +5.10 pp | yes | no | no | advantage-abstention margin fine sweep with measured train-source prior |
 | retrieval_residual_k4_mean_grid035040045_noopbonus003 | K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03 | complete | 35.42% | +5.68 pp | yes | no | no | scale-grid diagnostic for sparse mean-consensus residual transport |
 | retrieval_residual_k4_mean_grid035040045_srcscorebonus002 | K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, source-score bonus 0.02 | complete | 35.30% | +5.57 pp | yes | no | no | scale-grid diagnostic for measured-prior mean-consensus residual transport |
+| retrieval_residual_k4_mean_grid040045050_noopbonus003 | K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, no-op bonus 0.03 | complete | 35.36% | +5.62 pp | yes | no | no | upper tangent-length sweep for sparse mean-consensus residual transport |
+| retrieval_residual_k4_mean_grid040045050_srcscorebonus002 | K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, source-score bonus 0.02 | complete | 35.30% | +5.57 pp | yes | no | no | upper tangent-length sweep for measured-prior mean-consensus residual transport |
+| retrieval_residual_k4_mean_grid035045055_noopbonus003 | K4 mean-by-type residual retrieval, scales 0.35/0.45/0.55, margin 0.20, no-op bonus 0.03 | complete | 35.13% | +5.39 pp | yes | no | no | wide tangent-length sweep for sparse mean-consensus residual transport |
 | retrieval_residual_k4_mean_nooponly_noopbonus003 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, no-op bonus 0.03 | complete | 35.19% | +5.45 pp | yes | no | no | no-op-only residual-family ablation for sparse tangent transport |
 | retrieval_residual_k4_mean_nooponly_srcscorebonus002 | K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, source-score bonus 0.02 | complete | 35.19% | +5.45 pp | yes | no | no | no-op-only residual-family ablation with measured train-source prior |
 | 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 |

scripts/build_paper_analysis.py

@@ -211,6 +211,30 @@ METHODS = [
             "k4_grid035040045_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json"
         ),
     ),
+    MethodSpec(
+        key="residual_k4_consensus_grid040045050_noopbonus003",
+        label="K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, no-op bonus 0.03",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_"
+            "k4_grid040045050_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json"
+        ),
+    ),
+    MethodSpec(
+        key="residual_k4_consensus_grid040045050_srcscorebonus002",
+        label="K4 mean-by-type tangent consensus, scales 0.40/0.45/0.50, source-score bonus 0.02",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_"
+            "k4_grid040045050_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json"
+        ),
+    ),
+    MethodSpec(
+        key="residual_k4_consensus_grid035045055_noopbonus003",
+        label="K4 mean-by-type tangent consensus, scales 0.35/0.45/0.55, no-op bonus 0.03",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_"
+            "k4_grid035045055_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json"
+        ),
+    ),
     MethodSpec(
         key="residual_k4_consensus_nooponly_noopbonus003",
         label="K4 mean-by-type tangent consensus, no-op-only residuals, no-op bonus 0.03",

scripts/build_paper_table_status.py

@@ -525,6 +525,36 @@ SPECS = [
         story_role="scale-grid diagnostic for measured-prior mean-consensus residual transport",
         pending_job="14897989/14897991",
     ),
+    ResultSpec(
+        key="retrieval_residual_k4_mean_grid040045050_noopbonus003",
+        label="K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, no-op bonus 0.03",
+        path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="upper tangent-length sweep for sparse mean-consensus residual transport",
+        pending_job="14898107/14898110",
+    ),
+    ResultSpec(
+        key="retrieval_residual_k4_mean_grid040045050_srcscorebonus002",
+        label="K4 mean-by-type residual retrieval, scales 0.40/0.45/0.50, margin 0.20, source-score bonus 0.02",
+        path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid040045050_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="upper tangent-length sweep for measured-prior mean-consensus residual transport",
+        pending_job="14898108/14898111",
+    ),
+    ResultSpec(
+        key="retrieval_residual_k4_mean_grid035045055_noopbonus003",
+        label="K4 mean-by-type residual retrieval, scales 0.35/0.45/0.55, margin 0.20, no-op bonus 0.03",
+        path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_grid035045055_safe_margin0p20_mean_by_type_noopbonus0p03_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="wide tangent-length sweep for sparse mean-consensus residual transport",
+        pending_job="14898109/14898112",
+    ),
     ResultSpec(
         key="retrieval_residual_k4_mean_nooponly_noopbonus003",
         label="K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, no-op bonus 0.03",