anhtld commited on
Commit
5125e37
·
verified ·
1 Parent(s): 1bbdb12

Auto-sync: 2026-06-29 02:06:58

Browse files
results/paper_analysis.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "best_clean_key": "residual_k4_consensus_noopbonus003",
3
- "generated_utc": "2026-06-29T05:52:54+00:00",
4
  "mechanism_gap": {
5
  "best_clean_vs_direct_same_ckpt": 0.06956521739130428,
6
  "best_clean_vs_h16": 0.05507246376811592,
@@ -1350,6 +1350,148 @@
1350
  "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_noopbonus0p08_summary.json",
1351
  "std_success": 0.012049049096131323
1352
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1353
  "residual_k4_consensus_srcprogbonus003": {
1354
  "ci95_success": 0.03518915067855206,
1355
  "label": "K4 mean-by-type tangent consensus, source-progress bonus 0.03",
 
1
  {
2
  "best_clean_key": "residual_k4_consensus_noopbonus003",
3
+ "generated_utc": "2026-06-29T06:12:41+00:00",
4
  "mechanism_gap": {
5
  "best_clean_vs_direct_same_ckpt": 0.06956521739130428,
6
  "best_clean_vs_h16": 0.05507246376811592,
 
1350
  "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_noopbonus0p08_summary.json",
1351
  "std_success": 0.012049049096131323
1352
  },
1353
+ "residual_k4_consensus_nooponly_noopbonus003": {
1354
+ "ci95_success": 0.029409635829182355,
1355
+ "label": "K4 mean-by-type tangent consensus, no-op-only residuals, no-op bonus 0.03",
1356
+ "mean_action_mse_to_best": 0.39419553005419994,
1357
+ "mean_progress": 0.5656723719785794,
1358
+ "mean_success": 0.35188405797101446,
1359
+ "num_completed": 3,
1360
+ "per_task_success": {
1361
+ "LiftPegUpright-v1": {
1362
+ "mean_num_groups": 102.0,
1363
+ "mean_success": 0.26340865087329823,
1364
+ "std_success": 0.06057881385012102
1365
+ },
1366
+ "PickCube-v1": {
1367
+ "mean_num_groups": 196.66666666666666,
1368
+ "mean_success": 0.3244027510331859,
1369
+ "std_success": 0.026822320875242476
1370
+ },
1371
+ "PullCube-v1": {
1372
+ "mean_num_groups": 81.0,
1373
+ "mean_success": 0.20103794840636946,
1374
+ "std_success": 0.014667153114049147
1375
+ },
1376
+ "PushCube-v1": {
1377
+ "mean_num_groups": 102.0,
1378
+ "mean_success": 0.7414966050946671,
1379
+ "std_success": 0.06647794018181888
1380
+ },
1381
+ "StackCube-v1": {
1382
+ "mean_num_groups": 93.33333333333333,
1383
+ "mean_success": 0.20481740481740482,
1384
+ "std_success": 0.06861088450109236
1385
+ }
1386
+ },
1387
+ "seed_action_mse_to_best": {
1388
+ "0": 0.38096396585845427,
1389
+ "1": 0.38766786168612866,
1390
+ "2": 0.41395476261801695
1391
+ },
1392
+ "seed_progress": {
1393
+ "0": 0.5525477739123871,
1394
+ "1": 0.5665045300385226,
1395
+ "2": 0.5779648119848286
1396
+ },
1397
+ "seed_success": {
1398
+ "0": 0.34782608695652173,
1399
+ "1": 0.3426086956521739,
1400
+ "2": 0.3652173913043478
1401
+ },
1402
+ "selected_candidate_type_counts": {
1403
+ "retrieval_residual_policy_residual": 1671,
1404
+ "retrieval_residual_residual_no_op": 54
1405
+ },
1406
+ "selected_residual_scale_counts": {
1407
+ "0.4": 1725
1408
+ },
1409
+ "selected_type_outcomes": {
1410
+ "retrieval_residual_policy_residual": {
1411
+ "count": 1671.0,
1412
+ "mean_progress": 0.5605548383390697,
1413
+ "success_rate": 0.3464991023339318
1414
+ },
1415
+ "retrieval_residual_residual_no_op": {
1416
+ "count": 54.0,
1417
+ "mean_progress": 0.7240316073789641,
1418
+ "success_rate": 0.5185185185185185
1419
+ }
1420
+ },
1421
+ "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_nooponly_noopbonus0p03_summary.json",
1422
+ "std_success": 0.011838016148789656
1423
+ },
1424
+ "residual_k4_consensus_nooponly_srcscorebonus002": {
1425
+ "ci95_success": 0.029409635829182355,
1426
+ "label": "K4 mean-by-type tangent consensus, no-op-only residuals, source-score bonus 0.02",
1427
+ "mean_action_mse_to_best": 0.39407866501532823,
1428
+ "mean_progress": 0.5655461518514168,
1429
+ "mean_success": 0.35188405797101446,
1430
+ "num_completed": 3,
1431
+ "per_task_success": {
1432
+ "LiftPegUpright-v1": {
1433
+ "mean_num_groups": 102.0,
1434
+ "mean_success": 0.26340865087329823,
1435
+ "std_success": 0.06057881385012102
1436
+ },
1437
+ "PickCube-v1": {
1438
+ "mean_num_groups": 196.66666666666666,
1439
+ "mean_success": 0.3244027510331859,
1440
+ "std_success": 0.026822320875242476
1441
+ },
1442
+ "PullCube-v1": {
1443
+ "mean_num_groups": 81.0,
1444
+ "mean_success": 0.20103794840636946,
1445
+ "std_success": 0.014667153114049147
1446
+ },
1447
+ "PushCube-v1": {
1448
+ "mean_num_groups": 102.0,
1449
+ "mean_success": 0.7414966050946671,
1450
+ "std_success": 0.06647794018181888
1451
+ },
1452
+ "StackCube-v1": {
1453
+ "mean_num_groups": 93.33333333333333,
1454
+ "mean_success": 0.20481740481740482,
1455
+ "std_success": 0.06861088450109236
1456
+ }
1457
+ },
1458
+ "seed_action_mse_to_best": {
1459
+ "0": 0.38061467294783696,
1460
+ "1": 0.38757041037244644,
1461
+ "2": 0.4140509117257012
1462
+ },
1463
+ "seed_progress": {
1464
+ "0": 0.5521566874882126,
1465
+ "1": 0.5664698526813932,
1466
+ "2": 0.5780119153846444
1467
+ },
1468
+ "seed_success": {
1469
+ "0": 0.34782608695652173,
1470
+ "1": 0.3426086956521739,
1471
+ "2": 0.3652173913043478
1472
+ },
1473
+ "selected_candidate_type_counts": {
1474
+ "retrieval_residual_policy_residual": 1676,
1475
+ "retrieval_residual_residual_no_op": 49
1476
+ },
1477
+ "selected_residual_scale_counts": {
1478
+ "0.4": 1725
1479
+ },
1480
+ "selected_type_outcomes": {
1481
+ "retrieval_residual_policy_residual": {
1482
+ "count": 1676.0,
1483
+ "mean_progress": 0.5597072017399298,
1484
+ "success_rate": 0.3454653937947494
1485
+ },
1486
+ "retrieval_residual_residual_no_op": {
1487
+ "count": 49.0,
1488
+ "mean_progress": 0.7652620781137018,
1489
+ "success_rate": 0.5714285714285714
1490
+ }
1491
+ },
1492
+ "source": "results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_nooponly_srcscorebonus0p02_summary.json",
1493
+ "std_success": 0.011838016148789656
1494
+ },
1495
  "residual_k4_consensus_srcprogbonus003": {
1496
  "ci95_success": 0.03518915067855206,
1497
  "label": "K4 mean-by-type tangent consensus, source-progress bonus 0.03",
results/paper_analysis.md CHANGED
@@ -1,6 +1,6 @@
1
  # Paper Analysis
2
 
3
- Generated: `2026-06-29T05:52:54+00:00`
4
 
5
  ## Main Seed Statistics
6
 
@@ -24,6 +24,8 @@ Generated: `2026-06-29T05:52:54+00:00`
24
  | residual_k8_fieldsoftmax_grid_noopbonus003 | K8 field-softmax tangent transport, scales 0.35/0.40/0.45, no-op bonus 0.03 | 3 | 34.84% +/- 1.35 | +/- 3.36 | 56.55% | 0.397 | +5.10 pp |
25
  | residual_k4_consensus_noopbonus003 | K4 mean-by-type tangent consensus, no-op bonus 0.03 | 3 | 35.25% +/- 1.28 | +/- 3.18 | 56.68% | 0.395 | +5.51 pp |
26
  | 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 |
 
 
27
  | 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 |
28
  | 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 |
29
  | 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 |
 
1
  # Paper Analysis
2
 
3
+ Generated: `2026-06-29T06:12:41+00:00`
4
 
5
  ## Main Seed Statistics
6
 
 
24
  | residual_k8_fieldsoftmax_grid_noopbonus003 | K8 field-softmax tangent transport, scales 0.35/0.40/0.45, no-op bonus 0.03 | 3 | 34.84% +/- 1.35 | +/- 3.36 | 56.55% | 0.397 | +5.10 pp |
25
  | residual_k4_consensus_noopbonus003 | K4 mean-by-type tangent consensus, no-op bonus 0.03 | 3 | 35.25% +/- 1.28 | +/- 3.18 | 56.68% | 0.395 | +5.51 pp |
26
  | 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 |
27
+ | 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 |
28
+ | 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 |
29
  | 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 |
30
  | 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 |
31
  | 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 |
results/paper_table_status.json CHANGED
@@ -859,6 +859,44 @@
859
  "best_config": null,
860
  "gain_vs_h16_policy": 0.05449275362318845
861
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
862
  {
863
  "key": "retrieval_residual_k4_mean_noopbonus003_srcprog050",
864
  "label": "K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op bonus 0.03, source progress >= 0.50",
 
859
  "best_config": null,
860
  "gain_vs_h16_policy": 0.05449275362318845
861
  },
862
+ {
863
+ "key": "retrieval_residual_k4_mean_nooponly_noopbonus003",
864
+ "label": "K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, no-op bonus 0.03",
865
+ "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_nooponly_noopbonus0p03_summary.json",
866
+ "clean_deployment": "yes",
867
+ "same_state_proposals": "no",
868
+ "expert_proposal": "no",
869
+ "story_role": "no-op-only residual-family ablation for sparse tangent transport",
870
+ "fallback_success": null,
871
+ "pending_job": "14897563/14897565",
872
+ "path_exists": true,
873
+ "status": "complete",
874
+ "success": 0.35188405797101446,
875
+ "std_success": 0.011838016148789656,
876
+ "completed_seeds": null,
877
+ "num_completed": 3,
878
+ "best_config": null,
879
+ "gain_vs_h16_policy": 0.054492753623188395
880
+ },
881
+ {
882
+ "key": "retrieval_residual_k4_mean_nooponly_srcscorebonus002",
883
+ "label": "K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op-only residuals, source-score bonus 0.02",
884
+ "path": "h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_nooponly_srcscorebonus0p02_summary.json",
885
+ "clean_deployment": "yes",
886
+ "same_state_proposals": "no",
887
+ "expert_proposal": "no",
888
+ "story_role": "no-op-only residual-family ablation with measured train-source prior",
889
+ "fallback_success": null,
890
+ "pending_job": "14897564/14897566",
891
+ "path_exists": true,
892
+ "status": "complete",
893
+ "success": 0.35188405797101446,
894
+ "std_success": 0.011838016148789656,
895
+ "completed_seeds": null,
896
+ "num_completed": 3,
897
+ "best_config": null,
898
+ "gain_vs_h16_policy": 0.054492753623188395
899
+ },
900
  {
901
  "key": "retrieval_residual_k4_mean_noopbonus003_srcprog050",
902
  "label": "K4 mean-by-type residual retrieval, scale 0.40, margin 0.20, no-op bonus 0.03, source progress >= 0.50",