anhtld commited on
Commit
e1352f5
·
verified ·
1 Parent(s): 9692c5b

Auto-sync: 2026-06-29 07:52:27

Browse files
scripts/build_paper_analysis.py CHANGED
@@ -315,6 +315,38 @@ METHODS = [
315
  "k4_grid035040045_safe_margin0p20_mean_by_type_noopbonus0p03_consensus0p10_summary.json"
316
  ),
317
  ),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
318
  MethodSpec(
319
  key="residual_k4_consensus_grid035040045_noopbonus003_l2penalty005",
320
  label="K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03, action L2 penalty 0.05",
 
315
  "k4_grid035040045_safe_margin0p20_mean_by_type_noopbonus0p03_consensus0p10_summary.json"
316
  ),
317
  ),
318
+ MethodSpec(
319
+ key="repair_nearmiss_k4_grid025035050_margin020",
320
+ label="K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20",
321
+ summary_path=(
322
+ "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_"
323
+ "nearmiss_k4_grid025035050_margin0p20_summary.json"
324
+ ),
325
+ ),
326
+ MethodSpec(
327
+ key="repair_nearmiss_k4_grid035050075_margin020",
328
+ label="K4 near-miss-to-expert repair tangent, scales 0.35/0.50/0.75, margin 0.20",
329
+ summary_path=(
330
+ "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_"
331
+ "nearmiss_k4_grid035050075_margin0p20_summary.json"
332
+ ),
333
+ ),
334
+ MethodSpec(
335
+ key="repair_nearmiss_k4_grid025035050_margin010",
336
+ label="K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.10",
337
+ summary_path=(
338
+ "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_"
339
+ "nearmiss_k4_grid025035050_margin0p10_summary.json"
340
+ ),
341
+ ),
342
+ MethodSpec(
343
+ key="repair_safe_k4_grid025035050_margin020",
344
+ label="K4 safe-family-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20",
345
+ summary_path=(
346
+ "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_"
347
+ "safe_k4_grid025035050_margin0p20_summary.json"
348
+ ),
349
+ ),
350
  MethodSpec(
351
  key="residual_k4_consensus_grid035040045_noopbonus003_l2penalty005",
352
  label="K4 mean-by-type tangent consensus, scales 0.35/0.40/0.45, no-op bonus 0.03, action L2 penalty 0.05",
scripts/build_paper_table_status.py CHANGED
@@ -655,6 +655,46 @@ SPECS = [
655
  story_role="train-neighbor tangent-consensus confidence on the current best typed prior",
656
  pending_job="14903390/14903391",
657
  ),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
658
  ResultSpec(
659
  key="retrieval_residual_k4_mean_grid035040045_noopbonus003_l2penalty005",
660
  label="K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03, action L2 penalty 0.05",
 
655
  story_role="train-neighbor tangent-consensus confidence on the current best typed prior",
656
  pending_job="14903390/14903391",
657
  ),
658
+ ResultSpec(
659
+ key="retrieval_repair_nearmiss_k4_grid025035050_margin020",
660
+ label="K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20",
661
+ path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_nearmiss_k4_grid025035050_margin0p20_summary.json",
662
+ clean_deployment="yes",
663
+ same_state_proposals="no",
664
+ expert_proposal="no",
665
+ story_role="deployment-clean corrective tangent transport from train near-misses back toward expert actions",
666
+ pending_job="14904737/14904738",
667
+ ),
668
+ ResultSpec(
669
+ key="retrieval_repair_nearmiss_k4_grid035050075_margin020",
670
+ label="K4 near-miss-to-expert repair tangent, scales 0.35/0.50/0.75, margin 0.20",
671
+ path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_nearmiss_k4_grid035050075_margin0p20_summary.json",
672
+ clean_deployment="yes",
673
+ same_state_proposals="no",
674
+ expert_proposal="no",
675
+ story_role="repair-tangent scale diagnostic for near-miss counterfactual geometry",
676
+ pending_job="14904740/14904741",
677
+ ),
678
+ ResultSpec(
679
+ key="retrieval_repair_nearmiss_k4_grid025035050_margin010",
680
+ label="K4 near-miss-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.10",
681
+ path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_nearmiss_k4_grid025035050_margin0p10_summary.json",
682
+ clean_deployment="yes",
683
+ same_state_proposals="no",
684
+ expert_proposal="no",
685
+ story_role="repair-tangent abstention diagnostic for near-miss counterfactual geometry",
686
+ pending_job="14904742/14904743",
687
+ ),
688
+ ResultSpec(
689
+ key="retrieval_repair_safe_k4_grid025035050_margin020",
690
+ label="K4 safe-family-to-expert repair tangent, scales 0.25/0.35/0.50, margin 0.20",
691
+ path="h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_repair_safe_k4_grid025035050_margin0p20_summary.json",
692
+ clean_deployment="yes",
693
+ same_state_proposals="no",
694
+ expert_proposal="no",
695
+ story_role="repair-tangent family diagnostic including near-miss, no-op, and wrong-gripper corrections",
696
+ pending_job="14904744/14904745",
697
+ ),
698
  ResultSpec(
699
  key="retrieval_residual_k4_mean_grid035040045_noopbonus003_l2penalty005",
700
  label="K4 mean-by-type residual retrieval, scales 0.35/0.40/0.45, margin 0.20, no-op bonus 0.03, action L2 penalty 0.05",