Add ME-Dex 1.5 evaluation logs (Li Auto.)

#9
by Aroid721 - opened
This view is limited to 50 files because it contains too many changes. See the raw diff here.
Files changed (50) hide show
  1. me-dex-1-5/README.md +48 -0
  2. me-dex-1-5/leaderboard_me-dex-1-5.json +2456 -0
  3. me-dex-1-5/logo.png +3 -0
  4. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Equi/per_episode.jsonl +0 -0
  5. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Equi/summary.json +208 -0
  6. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Full/per_episode.jsonl +0 -0
  7. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Full/summary.json +199 -0
  8. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Inv/per_episode.jsonl +0 -0
  9. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Inv/summary.json +205 -0
  10. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/None/per_episode.jsonl +0 -0
  11. me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/None/summary.json +208 -0
  12. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Equi/per_episode.jsonl +0 -0
  13. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Equi/summary.json +265 -0
  14. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Full/per_episode.jsonl +0 -0
  15. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Full/summary.json +254 -0
  16. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Inv/per_episode.jsonl +0 -0
  17. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Inv/summary.json +260 -0
  18. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/None/per_episode.jsonl +0 -0
  19. me-dex-1-5/per_task_eval/06_fruit_bowl_loading/None/summary.json +260 -0
  20. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Equi/per_episode.jsonl +0 -0
  21. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Equi/summary.json +277 -0
  22. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Full/per_episode.jsonl +0 -0
  23. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Full/summary.json +261 -0
  24. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Inv/per_episode.jsonl +0 -0
  25. me-dex-1-5/per_task_eval/07_citrus_plate_loading/Inv/summary.json +273 -0
  26. me-dex-1-5/per_task_eval/07_citrus_plate_loading/None/per_episode.jsonl +0 -0
  27. me-dex-1-5/per_task_eval/07_citrus_plate_loading/None/summary.json +275 -0
  28. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Equi/per_episode.jsonl +0 -0
  29. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Equi/summary.json +240 -0
  30. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Full/per_episode.jsonl +0 -0
  31. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Full/summary.json +241 -0
  32. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Inv/per_episode.jsonl +0 -0
  33. me-dex-1-5/per_task_eval/08_frypan_stand_pour/Inv/summary.json +239 -0
  34. me-dex-1-5/per_task_eval/08_frypan_stand_pour/None/per_episode.jsonl +0 -0
  35. me-dex-1-5/per_task_eval/08_frypan_stand_pour/None/summary.json +240 -0
  36. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Equi/per_episode.jsonl +0 -0
  37. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Equi/summary.json +229 -0
  38. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Full/per_episode.jsonl +0 -0
  39. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Full/summary.json +237 -0
  40. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Inv/per_episode.jsonl +0 -0
  41. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Inv/summary.json +238 -0
  42. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/None/per_episode.jsonl +0 -0
  43. me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/None/summary.json +225 -0
  44. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Equi/per_episode.jsonl +0 -0
  45. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Equi/summary.json +260 -0
  46. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Full/per_episode.jsonl +0 -0
  47. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Full/summary.json +259 -0
  48. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Inv/per_episode.jsonl +0 -0
  49. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Inv/summary.json +263 -0
  50. me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/None/per_episode.jsonl +0 -0
me-dex-1-5/README.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ME-Dex 1.5
2
+
3
+ ME-Dex 1.5 is a tactile-aware world action model that jointly predicts video, tactile signals, and actions. Building on the link between contact forces and robot motion reflected in tactile signals, it combines autoregressive temporal modeling with flow matching to model visual, tactile, and action representations, allowing tactile information to guide action generation.
4
+
5
+ We replayed Bench2Dex demonstrations in simulation to collect three-axis tactile observations for training. We trained a single model across all 26 tasks for 60,000 steps, followed by 20,000 steps of fine-tuning for each task.
6
+
7
+ During evaluation, the policy receives RGB camera views, proprioception, and the instruction provided in the official scene description. To match the official evaluation protocol, we set force inputs to zero while keeping the tactile expert active.
8
+
9
+ The project page and code will be available soon.
10
+
11
+ ## Bench2Dex results
12
+
13
+ We evaluated the task-specific checkpoints on [the official Bench2Dex revision `c269f24b`](https://github.com/Bench2Dex/Bench2Dex/tree/c269f24bd39bd973e157a16c3c9c666cf5f43efc).
14
+
15
+ | Tasks | Episodes | Stable successes | SR | LSCR | Eff. | Score |
16
+ | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
17
+ | 26 | 5,200 | 3,047 | 58.5962% | 76.3125% | 0.979473 | 70.5036 |
18
+
19
+ ### Per-task results
20
+
21
+ | ID | Task | Stable successes / episodes | SR | LSCR | Eff. | Score |
22
+ | ---: | --- | ---: | ---: | ---: | ---: | ---: |
23
+ | 03 | Wine Glass Plate Balance | 90 / 200 | 45.0000% | 87.8333% | 1.048083 | 70.2558 |
24
+ | 06 | Fruit Bowl Loading | 174 / 200 | 87.0000% | 97.8750% | 0.930058 | 92.4943 |
25
+ | 07 | Citrus Plate Loading | 183 / 200 | 91.5000% | 97.5000% | 0.837078 | 93.4208 |
26
+ | 08 | Frypan Stand & Pour | 188 / 200 | 94.0000% | 96.2500% | 1.313465 | 98.7471 |
27
+ | 09 | Cleaner Box Loading | 145 / 200 | 72.5000% | 93.7500% | 0.909235 | 83.9049 |
28
+ | 12 | Screwdriver Box & Hammer | 132 / 200 | 66.0000% | 80.5000% | 0.980098 | 75.7260 |
29
+ | 21 | Condiment Box Loading | 157 / 200 | 78.5000% | 95.0000% | 0.928141 | 87.3564 |
30
+ | 22 | Tool Box Loading | 142 / 200 | 71.0000% | 82.6250% | 0.941639 | 78.5476 |
31
+ | 24 | Stationery Category Sorting | 61 / 200 | 30.5000% | 58.5000% | 0.999484 | 50.0448 |
32
+ | 26 | Canned Food Tray Arrangement | 111 / 200 | 55.5000% | 75.7500% | 0.931362 | 68.3761 |
33
+ | 27 | Ball Box Loading | 97 / 200 | 48.5000% | 62.7500% | 0.920494 | 59.2674 |
34
+ | 32 | Baking Tray Prep | 79 / 200 | 39.5000% | 54.7500% | 0.970917 | 52.1217 |
35
+ | 34 | Fridge Wine Interhand Pour | 99 / 200 | 49.5000% | 76.0000% | 1.012193 | 66.5969 |
36
+ | 42 | Trash Disposal | 105 / 200 | 52.5000% | 85.5000% | 0.938799 | 71.4880 |
37
+ | 43 | Fridge Fruit Shelf Sorting | 89 / 200 | 44.5000% | 63.0000% | 0.971438 | 58.0894 |
38
+ | 44 | Microwave Bowl Loading | 181 / 200 | 90.5000% | 95.0000% | 1.018788 | 93.6629 |
39
+ | 51 | Toilet Lid Cleaner Pour | 158 / 200 | 79.0000% | 84.0000% | 1.106688 | 84.4169 |
40
+ | 60 | Breadbasket Fast-Food Loading | 133 / 200 | 66.5000% | 88.6250% | 0.946621 | 79.2725 |
41
+ | 61 | Medicine Shoebox Pack | 125 / 200 | 62.5000% | 75.6667% | 1.051985 | 72.6949 |
42
+ | 62 | Shoebox Accessory Pack | 154 / 200 | 77.0000% | 83.1667% | 0.970526 | 81.7803 |
43
+ | 64 | Sports Ball Cup Sort | 75 / 200 | 37.5000% | 67.8750% | 0.772627 | 55.1450 |
44
+ | 67 | Faucet Cup Water Fill | 146 / 200 | 73.0000% | 83.6250% | 0.892604 | 79.4073 |
45
+ | 73 | Jigsaw Puzzle Assembly | 1 / 200 | 0.5000% | 50.0000% | 1.114890 | 33.8739 |
46
+ | 76 | Soup Serving | 25 / 200 | 12.5000% | 41.5000% | 0.973481 | 34.0348 |
47
+ | 79 | Bimanual Piano Melody | 64 / 200 | 32.0000% | 35.2500% | 0.937704 | 39.6395 |
48
+ | 80 | Gaming Desk Setup | 133 / 200 | 66.5000% | 71.8333% | 1.047899 | 72.7290 |
me-dex-1-5/leaderboard_me-dex-1-5.json ADDED
@@ -0,0 +1,2456 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": {
3
+ "id": "me-dex-1-5",
4
+ "label": "ME-Dex 1.5",
5
+ "variantLabel": "",
6
+ "family": "ME-Dex",
7
+ "modelVersion": "1.5",
8
+ "checkpoint": "Per-task step-20,000 checkpoints initialized from the shared step-60,000 model",
9
+ "inputs": "Four RGB views, proprioception and official scene language. Physical force is fixed at zero before tactile normalization and AE encoding.",
10
+ "protocolId": "standard",
11
+ "displayOrder": 141,
12
+ "trainingSetting": "60,000 mixed-task updates followed by 20,000 updates per task",
13
+ "resultVersion": "Official scene evaluation, 2026-11-19",
14
+ "updatedAt": "2026-11-19",
15
+ "evaluationSetting": "Task-specific checkpoint selection across 26 tasks and four channels with 50 episodes per task/channel.",
16
+ "sourceLabel": "Evaluation logs",
17
+ "sourceUrl": null,
18
+ "resultDetailsUrl": null,
19
+ "contributor": "Li Auto.",
20
+ "contributorUrl": null,
21
+ "type": "Visuo-tactile action model",
22
+ "modelUrl": null,
23
+ "logoPath": "leaderboard/me-dex-1-5/logo.png",
24
+ "logoAlt": null,
25
+ "weightsUrl": null,
26
+ "codeUrl": null
27
+ },
28
+ "overall": {
29
+ "taskCount": 26,
30
+ "scope": "task-macro",
31
+ "status": "reported",
32
+ "sr": 58.59615384615385,
33
+ "lscr": 76.3125,
34
+ "efficiency": 0.979472924297274,
35
+ "efficiencyStatus": "reported",
36
+ "efficiencyTaskCount": 26,
37
+ "channelSr": {
38
+ "None": 65.3076923076923,
39
+ "Equi.": 55.76923076923077,
40
+ "Inv.": 57.23076923076923,
41
+ "Full": 56.07692307692308
42
+ },
43
+ "channelLscr": {
44
+ "None": 80.57179487179486,
45
+ "Equi.": 73.72307692307692,
46
+ "Inv.": 76.03717948717949,
47
+ "Full": 74.91794871794872
48
+ },
49
+ "channelTaskCounts": {
50
+ "None": 26,
51
+ "Equi.": 26,
52
+ "Inv.": 26,
53
+ "Full": 26
54
+ }
55
+ },
56
+ "per_task": {
57
+ "12": {
58
+ "channels": {
59
+ "None": {
60
+ "status": "reported",
61
+ "episodes": 50,
62
+ "successes": 32,
63
+ "lscr": 79.0,
64
+ "source": {
65
+ "runId": "medex15-official-scene-20261118-12-none",
66
+ "perEpisodeFile": "per_task_eval/12_screwdriver_box_and_hammer/None/per_episode.jsonl",
67
+ "perEpisodeSha256": "5269a0ea408fa003dc2ba41623001145822dd364f89a3ce7f22e3afc2637254e",
68
+ "summaryFile": "per_task_eval/12_screwdriver_box_and_hammer/None/summary.json",
69
+ "summarySha256": "fbee417cdd921cd532b89fab4a3812631ca84b44429ff09630bb3770d2e70c58"
70
+ }
71
+ },
72
+ "Equi.": {
73
+ "status": "reported",
74
+ "episodes": 50,
75
+ "successes": 29,
76
+ "lscr": 78.0,
77
+ "source": {
78
+ "runId": "medex15-official-scene-20261118-12-cov_only",
79
+ "perEpisodeFile": "per_task_eval/12_screwdriver_box_and_hammer/Equi/per_episode.jsonl",
80
+ "perEpisodeSha256": "440b78f8bc46b39926b1e6d62838c2ca4b43b55b658c592f40fd5f777597ed4d",
81
+ "summaryFile": "per_task_eval/12_screwdriver_box_and_hammer/Equi/summary.json",
82
+ "summarySha256": "57dc59618027ccdc76c6992b3ebdaf261c07485d41c817f57900536a0a2e0394"
83
+ }
84
+ },
85
+ "Inv.": {
86
+ "status": "reported",
87
+ "episodes": 50,
88
+ "successes": 34,
89
+ "lscr": 81.5,
90
+ "source": {
91
+ "runId": "medex15-official-scene-20261118-12-inv_only",
92
+ "perEpisodeFile": "per_task_eval/12_screwdriver_box_and_hammer/Inv/per_episode.jsonl",
93
+ "perEpisodeSha256": "70d478b203e81a6c3bc08abe8062077482047184cecdd644f0c383a2da3ac8fb",
94
+ "summaryFile": "per_task_eval/12_screwdriver_box_and_hammer/Inv/summary.json",
95
+ "summarySha256": "bc68b6c3e016ecee16ee1079c4cd015dbc3b03326224291d5f7fdc58331bd847"
96
+ }
97
+ },
98
+ "Full": {
99
+ "status": "reported",
100
+ "episodes": 50,
101
+ "successes": 37,
102
+ "lscr": 83.5,
103
+ "source": {
104
+ "runId": "medex15-official-scene-20261118-12-inv_cov",
105
+ "perEpisodeFile": "per_task_eval/12_screwdriver_box_and_hammer/Full/per_episode.jsonl",
106
+ "perEpisodeSha256": "8c595f98f15e50af28a4a0bdc00dbaa4f57d970cb027b1853414f794417a5b25",
107
+ "summaryFile": "per_task_eval/12_screwdriver_box_and_hammer/Full/summary.json",
108
+ "summarySha256": "82b64450a81c7a714ffa5e583029d87413799bd008aed582f0effa1538f2db9c"
109
+ }
110
+ }
111
+ },
112
+ "mean": {
113
+ "status": "reported",
114
+ "sr": 66,
115
+ "srScope": "four-channel-mean",
116
+ "srEpisodes": 200,
117
+ "lscr": 80.5,
118
+ "lscrScope": "four-channel-mean",
119
+ "lscrEpisodes": 200,
120
+ "efficiency": 0.9800976852285841,
121
+ "efficiencyStatus": "reported"
122
+ }
123
+ },
124
+ "21": {
125
+ "channels": {
126
+ "None": {
127
+ "status": "reported",
128
+ "episodes": 50,
129
+ "successes": 44,
130
+ "lscr": 99.33333333333333,
131
+ "source": {
132
+ "runId": "medex15-official-scene-20261118-21-none",
133
+ "perEpisodeFile": "per_task_eval/21_condiment_box_loading/None/per_episode.jsonl",
134
+ "perEpisodeSha256": "31408d41f6aa241ea3005f93323bb65d45a58b5611f445184d777d18584d8be7",
135
+ "summaryFile": "per_task_eval/21_condiment_box_loading/None/summary.json",
136
+ "summarySha256": "7b885988902d7279e67f81b97c837fc5efcb7cb3b54b3e26c91ddf8c0a7e89c5"
137
+ }
138
+ },
139
+ "Equi.": {
140
+ "status": "reported",
141
+ "episodes": 50,
142
+ "successes": 38,
143
+ "lscr": 92.0,
144
+ "source": {
145
+ "runId": "medex15-official-scene-20261118-21-cov_only",
146
+ "perEpisodeFile": "per_task_eval/21_condiment_box_loading/Equi/per_episode.jsonl",
147
+ "perEpisodeSha256": "c5c6722cdf96ea65d760f34eb4c07b80faf952a90038e6d2242b0440a930372f",
148
+ "summaryFile": "per_task_eval/21_condiment_box_loading/Equi/summary.json",
149
+ "summarySha256": "51dbf519fa0c7092c9559e556c12511e2b5eec84ba9760b2d550099327d7f200"
150
+ }
151
+ },
152
+ "Inv.": {
153
+ "status": "reported",
154
+ "episodes": 50,
155
+ "successes": 41,
156
+ "lscr": 94.66666666666667,
157
+ "source": {
158
+ "runId": "medex15-official-scene-20261118-21-inv_only",
159
+ "perEpisodeFile": "per_task_eval/21_condiment_box_loading/Inv/per_episode.jsonl",
160
+ "perEpisodeSha256": "6331bfe284b47dcb91e05525694ab960134d7d33abe6a05de73396b38b130dd8",
161
+ "summaryFile": "per_task_eval/21_condiment_box_loading/Inv/summary.json",
162
+ "summarySha256": "c342919718cc4e21d7ab4195d3a2a0c4d8f7ac7aca43bad4c5bc66480b6962bf"
163
+ }
164
+ },
165
+ "Full": {
166
+ "status": "reported",
167
+ "episodes": 50,
168
+ "successes": 34,
169
+ "lscr": 94.0,
170
+ "source": {
171
+ "runId": "medex15-official-scene-20261118-21-inv_cov",
172
+ "perEpisodeFile": "per_task_eval/21_condiment_box_loading/Full/per_episode.jsonl",
173
+ "perEpisodeSha256": "04242e7916337626801176750ef8c9c08b29b7f7d6366ad5c0593530592b383e",
174
+ "summaryFile": "per_task_eval/21_condiment_box_loading/Full/summary.json",
175
+ "summarySha256": "93ccc5e3d9f0297f4646266ff69a311de2e71c4c32cb49d0dd80b72105d62f88"
176
+ }
177
+ }
178
+ },
179
+ "mean": {
180
+ "status": "reported",
181
+ "sr": 78.5,
182
+ "srScope": "four-channel-mean",
183
+ "srEpisodes": 200,
184
+ "lscr": 95.0,
185
+ "lscrScope": "four-channel-mean",
186
+ "lscrEpisodes": 200,
187
+ "efficiency": 0.928140737643102,
188
+ "efficiencyStatus": "reported"
189
+ }
190
+ },
191
+ "22": {
192
+ "channels": {
193
+ "None": {
194
+ "status": "reported",
195
+ "episodes": 50,
196
+ "successes": 42,
197
+ "lscr": 90.5,
198
+ "source": {
199
+ "runId": "medex15-official-scene-20261118-22-none",
200
+ "perEpisodeFile": "per_task_eval/22_tool_box_loading/None/per_episode.jsonl",
201
+ "perEpisodeSha256": "5487e3e3e6042fc1b536d283a2a40a1e1be0a4a0f3bec0573515c4c899e690f0",
202
+ "summaryFile": "per_task_eval/22_tool_box_loading/None/summary.json",
203
+ "summarySha256": "f054969e037ac4e547fdbcd7c03b7826f0da3cf6da7ab19a3cc90f678a40e0d9"
204
+ }
205
+ },
206
+ "Equi.": {
207
+ "status": "reported",
208
+ "episodes": 50,
209
+ "successes": 37,
210
+ "lscr": 86.5,
211
+ "source": {
212
+ "runId": "medex15-official-scene-20261118-22-cov_only",
213
+ "perEpisodeFile": "per_task_eval/22_tool_box_loading/Equi/per_episode.jsonl",
214
+ "perEpisodeSha256": "96db8f8a7d99d4ac1f11beeb7f39ac29c446e1fb7c16a0cf0bc66fea8bbc327f",
215
+ "summaryFile": "per_task_eval/22_tool_box_loading/Equi/summary.json",
216
+ "summarySha256": "94c6ab9c1837398cd7b2bad010b7a0dacedc17be2f821fd858c90f397367033c"
217
+ }
218
+ },
219
+ "Inv.": {
220
+ "status": "reported",
221
+ "episodes": 50,
222
+ "successes": 31,
223
+ "lscr": 75.5,
224
+ "source": {
225
+ "runId": "medex15-official-scene-20261118-22-inv_only",
226
+ "perEpisodeFile": "per_task_eval/22_tool_box_loading/Inv/per_episode.jsonl",
227
+ "perEpisodeSha256": "7e06d1e6590aed9d27132cffb5f08ede0b68773957118913bdf8cfc03a7b057d",
228
+ "summaryFile": "per_task_eval/22_tool_box_loading/Inv/summary.json",
229
+ "summarySha256": "aa4ddfba045ef3405b4855f29efc13c2cfef25f6fa1610856fa955bf6fa96d9f"
230
+ }
231
+ },
232
+ "Full": {
233
+ "status": "reported",
234
+ "episodes": 50,
235
+ "successes": 32,
236
+ "lscr": 78.0,
237
+ "source": {
238
+ "runId": "medex15-official-scene-20261118-22-inv_cov",
239
+ "perEpisodeFile": "per_task_eval/22_tool_box_loading/Full/per_episode.jsonl",
240
+ "perEpisodeSha256": "21e762b45657bbb63531549ab5e496ef9aa649ec39c172e1022cd6f4fd22da44",
241
+ "summaryFile": "per_task_eval/22_tool_box_loading/Full/summary.json",
242
+ "summarySha256": "424f956ff33322de380823ed42d0679bbc4e4dd21aee385fbbb48d62f2ea4883"
243
+ }
244
+ }
245
+ },
246
+ "mean": {
247
+ "status": "reported",
248
+ "sr": 71,
249
+ "srScope": "four-channel-mean",
250
+ "srEpisodes": 200,
251
+ "lscr": 82.625,
252
+ "lscrScope": "four-channel-mean",
253
+ "lscrEpisodes": 200,
254
+ "efficiency": 0.9416390567140909,
255
+ "efficiencyStatus": "reported"
256
+ }
257
+ },
258
+ "24": {
259
+ "channels": {
260
+ "None": {
261
+ "status": "reported",
262
+ "episodes": 50,
263
+ "successes": 18,
264
+ "lscr": 57.49999999999999,
265
+ "source": {
266
+ "runId": "medex15-official-scene-20261118-24-none",
267
+ "perEpisodeFile": "per_task_eval/24_stationery_category_sorting/None/per_episode.jsonl",
268
+ "perEpisodeSha256": "6262d310d13f5e3cab2f14e6b5edf97913330e8707c192c2c1ecb31b74933b7e",
269
+ "summaryFile": "per_task_eval/24_stationery_category_sorting/None/summary.json",
270
+ "summarySha256": "5813ea49ae1e00539d9ccc5d29144108c6fab8cd5210c76e681c37976f9e4178"
271
+ }
272
+ },
273
+ "Equi.": {
274
+ "status": "reported",
275
+ "episodes": 50,
276
+ "successes": 14,
277
+ "lscr": 56.49999999999999,
278
+ "source": {
279
+ "runId": "medex15-official-scene-20261118-24-cov_only",
280
+ "perEpisodeFile": "per_task_eval/24_stationery_category_sorting/Equi/per_episode.jsonl",
281
+ "perEpisodeSha256": "8d704587860f3fd43e06abf1abbb4d2c3a094bd4f4fb3325cb50572352f9dd4c",
282
+ "summaryFile": "per_task_eval/24_stationery_category_sorting/Equi/summary.json",
283
+ "summarySha256": "1543dee6a7bc984452fbf66521703f8195792eec669b7b8d8d7198f09a871b0b"
284
+ }
285
+ },
286
+ "Inv.": {
287
+ "status": "reported",
288
+ "episodes": 50,
289
+ "successes": 12,
290
+ "lscr": 59.0,
291
+ "source": {
292
+ "runId": "medex15-official-scene-20261118-24-inv_only",
293
+ "perEpisodeFile": "per_task_eval/24_stationery_category_sorting/Inv/per_episode.jsonl",
294
+ "perEpisodeSha256": "96373e85cf218d700622d009a45851e09d2023b722791547f2a9d4f99ea2379e",
295
+ "summaryFile": "per_task_eval/24_stationery_category_sorting/Inv/summary.json",
296
+ "summarySha256": "65076cc41902ed8248dcd176fdf3a4c2075e93f11bea11bfc363ea89226d4905"
297
+ }
298
+ },
299
+ "Full": {
300
+ "status": "reported",
301
+ "episodes": 50,
302
+ "successes": 17,
303
+ "lscr": 61.0,
304
+ "source": {
305
+ "runId": "medex15-official-scene-20261118-24-inv_cov",
306
+ "perEpisodeFile": "per_task_eval/24_stationery_category_sorting/Full/per_episode.jsonl",
307
+ "perEpisodeSha256": "2b289bf6b5f7adbe9d7438a39391e8cd60cfbe0f83dc51391265306757d57561",
308
+ "summaryFile": "per_task_eval/24_stationery_category_sorting/Full/summary.json",
309
+ "summarySha256": "23d9707f0c49c46615292339a6e563653cea943cfb5b8a24b861142d9a4989d2"
310
+ }
311
+ }
312
+ },
313
+ "mean": {
314
+ "status": "reported",
315
+ "sr": 30.5,
316
+ "srScope": "four-channel-mean",
317
+ "srEpisodes": 200,
318
+ "lscr": 58.5,
319
+ "lscrScope": "four-channel-mean",
320
+ "lscrEpisodes": 200,
321
+ "efficiency": 0.999484481292517,
322
+ "efficiencyStatus": "reported"
323
+ }
324
+ },
325
+ "26": {
326
+ "channels": {
327
+ "None": {
328
+ "status": "reported",
329
+ "episodes": 50,
330
+ "successes": 32,
331
+ "lscr": 88.5,
332
+ "source": {
333
+ "runId": "medex15-official-scene-20261118-26-none",
334
+ "perEpisodeFile": "per_task_eval/26_canned_food_tray_line_arrangement/None/per_episode.jsonl",
335
+ "perEpisodeSha256": "cd657d02667f3a1ab3f9fe3f2d352310e09e4b95f85874b369ade685e18a0e77",
336
+ "summaryFile": "per_task_eval/26_canned_food_tray_line_arrangement/None/summary.json",
337
+ "summarySha256": "63766888f8378e2f0741a16ae8ac464c9c50733fa893bcef13783fb38ecb86e3"
338
+ }
339
+ },
340
+ "Equi.": {
341
+ "status": "reported",
342
+ "episodes": 50,
343
+ "successes": 25,
344
+ "lscr": 65.0,
345
+ "source": {
346
+ "runId": "medex15-official-scene-20261118-26-cov_only",
347
+ "perEpisodeFile": "per_task_eval/26_canned_food_tray_line_arrangement/Equi/per_episode.jsonl",
348
+ "perEpisodeSha256": "9626906d873a05481ab9e5a1ace43456766926cf35b8369f2376ab56f6984745",
349
+ "summaryFile": "per_task_eval/26_canned_food_tray_line_arrangement/Equi/summary.json",
350
+ "summarySha256": "461e3033b23ffff5a5eacfbe224477e9330148f51d77137bfd01afcd8509cefd"
351
+ }
352
+ },
353
+ "Inv.": {
354
+ "status": "reported",
355
+ "episodes": 50,
356
+ "successes": 26,
357
+ "lscr": 74.0,
358
+ "source": {
359
+ "runId": "medex15-official-scene-20261118-26-inv_only",
360
+ "perEpisodeFile": "per_task_eval/26_canned_food_tray_line_arrangement/Inv/per_episode.jsonl",
361
+ "perEpisodeSha256": "ef2197aad19a8ff0101c2954b3c82ead751ddd4328dbe2eae79f1439bb758826",
362
+ "summaryFile": "per_task_eval/26_canned_food_tray_line_arrangement/Inv/summary.json",
363
+ "summarySha256": "099df81c8b11db83bcc247264afd9331408057d64173bddb46c9b68c1a96a52c"
364
+ }
365
+ },
366
+ "Full": {
367
+ "status": "reported",
368
+ "episodes": 50,
369
+ "successes": 28,
370
+ "lscr": 75.5,
371
+ "source": {
372
+ "runId": "medex15-official-scene-20261118-26-inv_cov",
373
+ "perEpisodeFile": "per_task_eval/26_canned_food_tray_line_arrangement/Full/per_episode.jsonl",
374
+ "perEpisodeSha256": "b98d1e891514de1f07c2d741497c108e1fbcfaefb5d333e21a7e7767cc01a864",
375
+ "summaryFile": "per_task_eval/26_canned_food_tray_line_arrangement/Full/summary.json",
376
+ "summarySha256": "cf13a9f0080cb79292fa8e632edb6808acb8b05aacec6587a84fc21109c9498d"
377
+ }
378
+ }
379
+ },
380
+ "mean": {
381
+ "status": "reported",
382
+ "sr": 55.5,
383
+ "srScope": "four-channel-mean",
384
+ "srEpisodes": 200,
385
+ "lscr": 75.75,
386
+ "lscrScope": "four-channel-mean",
387
+ "lscrEpisodes": 200,
388
+ "efficiency": 0.9313623489971677,
389
+ "efficiencyStatus": "reported"
390
+ }
391
+ },
392
+ "27": {
393
+ "channels": {
394
+ "None": {
395
+ "status": "reported",
396
+ "episodes": 50,
397
+ "successes": 30,
398
+ "lscr": 71.0,
399
+ "source": {
400
+ "runId": "medex15-official-scene-20261118-27-none",
401
+ "perEpisodeFile": "per_task_eval/27_ball_box_loading/None/per_episode.jsonl",
402
+ "perEpisodeSha256": "a36ce63c2f780710a38bb598d9ebe20d3955cbd1ed91d97300de8ebebfdf530f",
403
+ "summaryFile": "per_task_eval/27_ball_box_loading/None/summary.json",
404
+ "summarySha256": "c83b94e706e5ec7ed4ddc15a1e7bf5df78eed1b20ffdd2f9b32ffa133b377197"
405
+ }
406
+ },
407
+ "Equi.": {
408
+ "status": "reported",
409
+ "episodes": 50,
410
+ "successes": 17,
411
+ "lscr": 50.0,
412
+ "source": {
413
+ "runId": "medex15-official-scene-20261118-27-cov_only",
414
+ "perEpisodeFile": "per_task_eval/27_ball_box_loading/Equi/per_episode.jsonl",
415
+ "perEpisodeSha256": "c833afcd63723832d5756ec85452929c7bd1935ea1b0e3971fd5c580e8c45dea",
416
+ "summaryFile": "per_task_eval/27_ball_box_loading/Equi/summary.json",
417
+ "summarySha256": "a1ba37a9484085d491f22b37edbe10191897c9c8da04c31c2cbe342e62c4ea14"
418
+ }
419
+ },
420
+ "Inv.": {
421
+ "status": "reported",
422
+ "episodes": 50,
423
+ "successes": 29,
424
+ "lscr": 69.5,
425
+ "source": {
426
+ "runId": "medex15-official-scene-20261118-27-inv_only",
427
+ "perEpisodeFile": "per_task_eval/27_ball_box_loading/Inv/per_episode.jsonl",
428
+ "perEpisodeSha256": "cc8a258ccac7d7b589e53c03cc491f8f434cd3519e1a573d2ce50c79304afcc8",
429
+ "summaryFile": "per_task_eval/27_ball_box_loading/Inv/summary.json",
430
+ "summarySha256": "fd2793318a1f946225bfb28561f1ff1c8912337d0e585e8c1979079f13c67d43"
431
+ }
432
+ },
433
+ "Full": {
434
+ "status": "reported",
435
+ "episodes": 50,
436
+ "successes": 21,
437
+ "lscr": 60.5,
438
+ "source": {
439
+ "runId": "medex15-official-scene-20261118-27-inv_cov",
440
+ "perEpisodeFile": "per_task_eval/27_ball_box_loading/Full/per_episode.jsonl",
441
+ "perEpisodeSha256": "8d11eeb92ceb831dde34b456a7ec9b1e7007ad9261f0f7e65aac7aead16f4cf8",
442
+ "summaryFile": "per_task_eval/27_ball_box_loading/Full/summary.json",
443
+ "summarySha256": "8896a531a0914a89c8ea0e3b9f48318d23a6969cc0900f6c7f0098b65293183c"
444
+ }
445
+ }
446
+ },
447
+ "mean": {
448
+ "status": "reported",
449
+ "sr": 48.5,
450
+ "srScope": "four-channel-mean",
451
+ "srEpisodes": 200,
452
+ "lscr": 62.75,
453
+ "lscrScope": "four-channel-mean",
454
+ "lscrEpisodes": 200,
455
+ "efficiency": 0.9204944178628389,
456
+ "efficiencyStatus": "reported"
457
+ }
458
+ },
459
+ "32": {
460
+ "channels": {
461
+ "None": {
462
+ "status": "reported",
463
+ "episodes": 50,
464
+ "successes": 25,
465
+ "lscr": 65.0,
466
+ "source": {
467
+ "runId": "medex15-official-scene-20261118-32-none",
468
+ "perEpisodeFile": "per_task_eval/32_baking_tray_prep_with_tools/None/per_episode.jsonl",
469
+ "perEpisodeSha256": "aba8d70babe0b68b77b23ceb324d9f01e0494c603828e2d19ca37a7bf73c56fe",
470
+ "summaryFile": "per_task_eval/32_baking_tray_prep_with_tools/None/summary.json",
471
+ "summarySha256": "cd73d96b0589a34d509d211ea0707227de4c1721c19691c5c9ce756e57cd21c8"
472
+ }
473
+ },
474
+ "Equi.": {
475
+ "status": "reported",
476
+ "episodes": 50,
477
+ "successes": 15,
478
+ "lscr": 46.5,
479
+ "source": {
480
+ "runId": "medex15-official-scene-20261118-32-cov_only",
481
+ "perEpisodeFile": "per_task_eval/32_baking_tray_prep_with_tools/Equi/per_episode.jsonl",
482
+ "perEpisodeSha256": "77d0084bf73d376da20c4a31573e526a3cdfea18cef9281c085c4140f82aa3be",
483
+ "summaryFile": "per_task_eval/32_baking_tray_prep_with_tools/Equi/summary.json",
484
+ "summarySha256": "a83e8cdfc79303c1782e83d11acd6da0b0acb178192a4e32c683c04828763851"
485
+ }
486
+ },
487
+ "Inv.": {
488
+ "status": "reported",
489
+ "episodes": 50,
490
+ "successes": 20,
491
+ "lscr": 55.00000000000001,
492
+ "source": {
493
+ "runId": "medex15-official-scene-20261118-32-inv_only",
494
+ "perEpisodeFile": "per_task_eval/32_baking_tray_prep_with_tools/Inv/per_episode.jsonl",
495
+ "perEpisodeSha256": "f7bdb14a035b9d589e2a72bc82b4f8666b3bad19491bcdc6cf847d8c05e67737",
496
+ "summaryFile": "per_task_eval/32_baking_tray_prep_with_tools/Inv/summary.json",
497
+ "summarySha256": "7ffd569ff4ce476f22cb3cbc665f0c0f2a3f01ea9ac43b74485abec09e64b8df"
498
+ }
499
+ },
500
+ "Full": {
501
+ "status": "reported",
502
+ "episodes": 50,
503
+ "successes": 19,
504
+ "lscr": 52.5,
505
+ "source": {
506
+ "runId": "medex15-official-scene-20261118-32-inv_cov",
507
+ "perEpisodeFile": "per_task_eval/32_baking_tray_prep_with_tools/Full/per_episode.jsonl",
508
+ "perEpisodeSha256": "6da9522577fd7f4fe0c46755ed39d160b151b371dc0d4a3a25e1908fc5410c3c",
509
+ "summaryFile": "per_task_eval/32_baking_tray_prep_with_tools/Full/summary.json",
510
+ "summarySha256": "7b79ba53173d838508f1920d2a2efc02493cff1f927cb7d77794d2b4627a6bd7"
511
+ }
512
+ }
513
+ },
514
+ "mean": {
515
+ "status": "reported",
516
+ "sr": 39.5,
517
+ "srScope": "four-channel-mean",
518
+ "srEpisodes": 200,
519
+ "lscr": 54.75,
520
+ "lscrScope": "four-channel-mean",
521
+ "lscrEpisodes": 200,
522
+ "efficiency": 0.9709165963035932,
523
+ "efficiencyStatus": "reported"
524
+ }
525
+ },
526
+ "34": {
527
+ "channels": {
528
+ "None": {
529
+ "status": "reported",
530
+ "episodes": 50,
531
+ "successes": 30,
532
+ "lscr": 81.19999999999999,
533
+ "source": {
534
+ "runId": "medex15-official-scene-20261118-34-none",
535
+ "perEpisodeFile": "per_task_eval/34_fridge_wine_interhand_pour/None/per_episode.jsonl",
536
+ "perEpisodeSha256": "1e9fea8ebe3652bff2a1d71c5ca9fd4eb176c0a842343415ff89009883efb235",
537
+ "summaryFile": "per_task_eval/34_fridge_wine_interhand_pour/None/summary.json",
538
+ "summarySha256": "aeb4c6b2e5e4309689ee536cc065147f38a1d3fb119e6903655c6fa7c2bf0360"
539
+ }
540
+ },
541
+ "Equi.": {
542
+ "status": "reported",
543
+ "episodes": 50,
544
+ "successes": 20,
545
+ "lscr": 70.8,
546
+ "source": {
547
+ "runId": "medex15-official-scene-20261118-34-cov_only",
548
+ "perEpisodeFile": "per_task_eval/34_fridge_wine_interhand_pour/Equi/per_episode.jsonl",
549
+ "perEpisodeSha256": "809b4626d7f3194411b5bd850725ea7a6b12889d16616cc309781aae63054cfe",
550
+ "summaryFile": "per_task_eval/34_fridge_wine_interhand_pour/Equi/summary.json",
551
+ "summarySha256": "53a172faeaa20d205bd020c80becc32767b66a3d8b4679c6940a1aae819b904a"
552
+ }
553
+ },
554
+ "Inv.": {
555
+ "status": "reported",
556
+ "episodes": 50,
557
+ "successes": 29,
558
+ "lscr": 80.8,
559
+ "source": {
560
+ "runId": "medex15-official-scene-20261118-34-inv_only",
561
+ "perEpisodeFile": "per_task_eval/34_fridge_wine_interhand_pour/Inv/per_episode.jsonl",
562
+ "perEpisodeSha256": "71821f96f463064ec2ecac90f965231c0268073a85e61028b4e35bb297597f60",
563
+ "summaryFile": "per_task_eval/34_fridge_wine_interhand_pour/Inv/summary.json",
564
+ "summarySha256": "efe6149c4995cb4dfc189cdb32267aa02cc5f2a4c6822f8d80dee651190d3409"
565
+ }
566
+ },
567
+ "Full": {
568
+ "status": "reported",
569
+ "episodes": 50,
570
+ "successes": 20,
571
+ "lscr": 71.2,
572
+ "source": {
573
+ "runId": "medex15-official-scene-20261118-34-inv_cov",
574
+ "perEpisodeFile": "per_task_eval/34_fridge_wine_interhand_pour/Full/per_episode.jsonl",
575
+ "perEpisodeSha256": "e70bc9bdfc0aca8921b62173ac0d7193157d1322e60efd364b0e10637bf2db3f",
576
+ "summaryFile": "per_task_eval/34_fridge_wine_interhand_pour/Full/summary.json",
577
+ "summarySha256": "06c11d19aebb7b29cd2155098b549540cb29517ac32f0f3d7db338557add0ad7"
578
+ }
579
+ }
580
+ },
581
+ "mean": {
582
+ "status": "reported",
583
+ "sr": 49.5,
584
+ "srScope": "four-channel-mean",
585
+ "srEpisodes": 200,
586
+ "lscr": 76.0,
587
+ "lscrScope": "four-channel-mean",
588
+ "lscrEpisodes": 200,
589
+ "efficiency": 1.0121932747652227,
590
+ "efficiencyStatus": "reported"
591
+ }
592
+ },
593
+ "42": {
594
+ "channels": {
595
+ "None": {
596
+ "status": "reported",
597
+ "episodes": 50,
598
+ "successes": 28,
599
+ "lscr": 87.33333333333333,
600
+ "source": {
601
+ "runId": "medex15-official-scene-20261118-42-none",
602
+ "perEpisodeFile": "per_task_eval/42_trash_disposal/None/per_episode.jsonl",
603
+ "perEpisodeSha256": "21d615772f300bfd124435b6d6efb1d7f56877511703fc9bd98c4b0f5f63e2e9",
604
+ "summaryFile": "per_task_eval/42_trash_disposal/None/summary.json",
605
+ "summarySha256": "51d7cc71334899b33bd20be0b690c2ebdf87adb87778b3349b21cf6edf7e8eab"
606
+ }
607
+ },
608
+ "Equi.": {
609
+ "status": "reported",
610
+ "episodes": 50,
611
+ "successes": 27,
612
+ "lscr": 82.0,
613
+ "source": {
614
+ "runId": "medex15-official-scene-20261118-42-cov_only",
615
+ "perEpisodeFile": "per_task_eval/42_trash_disposal/Equi/per_episode.jsonl",
616
+ "perEpisodeSha256": "f6b351348de6a1d729f03b7a03c02b08daf2807cd97c75391b95bbc57c345da2",
617
+ "summaryFile": "per_task_eval/42_trash_disposal/Equi/summary.json",
618
+ "summarySha256": "0897d6d7fa46401995886b6ab2d0c0caa160a4e44065becd4aefc4ed64d1eb7b"
619
+ }
620
+ },
621
+ "Inv.": {
622
+ "status": "reported",
623
+ "episodes": 50,
624
+ "successes": 25,
625
+ "lscr": 83.33333333333334,
626
+ "source": {
627
+ "runId": "medex15-official-scene-20261118-42-inv_only",
628
+ "perEpisodeFile": "per_task_eval/42_trash_disposal/Inv/per_episode.jsonl",
629
+ "perEpisodeSha256": "bf40e887143b3dd6e21d8282b048342930d200f068b117ec965e26070dd8421f",
630
+ "summaryFile": "per_task_eval/42_trash_disposal/Inv/summary.json",
631
+ "summarySha256": "e4fb0e0939e6dfcacc5393f7480b111d9ee9202916bb738257595f763746fbf5"
632
+ }
633
+ },
634
+ "Full": {
635
+ "status": "reported",
636
+ "episodes": 50,
637
+ "successes": 25,
638
+ "lscr": 89.33333333333333,
639
+ "source": {
640
+ "runId": "medex15-official-scene-20261118-42-inv_cov",
641
+ "perEpisodeFile": "per_task_eval/42_trash_disposal/Full/per_episode.jsonl",
642
+ "perEpisodeSha256": "558d8548bf88e6da7b4274a5dcc47d50de2c56126e651de3006a34c2d5a8ea18",
643
+ "summaryFile": "per_task_eval/42_trash_disposal/Full/summary.json",
644
+ "summarySha256": "de70911cdea5552f52278dced445ee6eb767013c783e7db19189e9cd3b1c3d06"
645
+ }
646
+ }
647
+ },
648
+ "mean": {
649
+ "status": "reported",
650
+ "sr": 52.5,
651
+ "srScope": "four-channel-mean",
652
+ "srEpisodes": 200,
653
+ "lscr": 85.5,
654
+ "lscrScope": "four-channel-mean",
655
+ "lscrEpisodes": 200,
656
+ "efficiency": 0.9387987284590932,
657
+ "efficiencyStatus": "reported"
658
+ }
659
+ },
660
+ "43": {
661
+ "channels": {
662
+ "None": {
663
+ "status": "reported",
664
+ "episodes": 50,
665
+ "successes": 23,
666
+ "lscr": 61.5,
667
+ "source": {
668
+ "runId": "medex15-official-scene-20261118-43-none",
669
+ "perEpisodeFile": "per_task_eval/43_fridge_fruit_shelf_sorting/None/per_episode.jsonl",
670
+ "perEpisodeSha256": "9705b79cda52e24ebfbe3d85b6cc0d49965322ce6fa7eb286ccbc7533945a149",
671
+ "summaryFile": "per_task_eval/43_fridge_fruit_shelf_sorting/None/summary.json",
672
+ "summarySha256": "3f9179ef7ed4952a3f4985845ca5d5668b6e5e893a1e2d88c3475ed9cf8aca4b"
673
+ }
674
+ },
675
+ "Equi.": {
676
+ "status": "reported",
677
+ "episodes": 50,
678
+ "successes": 20,
679
+ "lscr": 56.99999999999999,
680
+ "source": {
681
+ "runId": "medex15-official-scene-20261118-43-cov_only",
682
+ "perEpisodeFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Equi/per_episode.jsonl",
683
+ "perEpisodeSha256": "9f3e83a25a3b9d71eb5282fbe0a867ac6233235213ac77d0d620fe310cfbf75d",
684
+ "summaryFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Equi/summary.json",
685
+ "summarySha256": "dc93b2175f31537512336879c45a0353a485429f3c1fd4233371b1fbcd3dda98"
686
+ }
687
+ },
688
+ "Inv.": {
689
+ "status": "reported",
690
+ "episodes": 50,
691
+ "successes": 24,
692
+ "lscr": 67.5,
693
+ "source": {
694
+ "runId": "medex15-official-scene-20261118-43-inv_only",
695
+ "perEpisodeFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Inv/per_episode.jsonl",
696
+ "perEpisodeSha256": "ed797bdd000b907073007fab767c4c30b3c23f406cf7f0c0290ce71ba3610f88",
697
+ "summaryFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Inv/summary.json",
698
+ "summarySha256": "c5219d3913cfc857de3e39e1c4c8179cb613c6939527c8fe9123a589081e7e97"
699
+ }
700
+ },
701
+ "Full": {
702
+ "status": "reported",
703
+ "episodes": 50,
704
+ "successes": 22,
705
+ "lscr": 66.0,
706
+ "source": {
707
+ "runId": "medex15-official-scene-20261118-43-inv_cov",
708
+ "perEpisodeFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Full/per_episode.jsonl",
709
+ "perEpisodeSha256": "d9a71fd7de9051c1f81918ef7b90b470e8be2c37f057c7676b904215cc909b70",
710
+ "summaryFile": "per_task_eval/43_fridge_fruit_shelf_sorting/Full/summary.json",
711
+ "summarySha256": "f530561b0b0894a8f479c4ab4102bc262b55acc0ce0f88ab9654bda9b979b5c9"
712
+ }
713
+ }
714
+ },
715
+ "mean": {
716
+ "status": "reported",
717
+ "sr": 44.5,
718
+ "srScope": "four-channel-mean",
719
+ "srEpisodes": 200,
720
+ "lscr": 63.0,
721
+ "lscrScope": "four-channel-mean",
722
+ "lscrEpisodes": 200,
723
+ "efficiency": 0.9714383240853198,
724
+ "efficiencyStatus": "reported"
725
+ }
726
+ },
727
+ "44": {
728
+ "channels": {
729
+ "None": {
730
+ "status": "reported",
731
+ "episodes": 50,
732
+ "successes": 47,
733
+ "lscr": 96.5,
734
+ "source": {
735
+ "runId": "medex15-official-scene-20261118-44-none",
736
+ "perEpisodeFile": "per_task_eval/44_microwave_bowl_loading/None/per_episode.jsonl",
737
+ "perEpisodeSha256": "0623cc26ba4d7773fa19b3dd1b3d2a2ed6cd9e487e31f5386b3d86e81ae3f54c",
738
+ "summaryFile": "per_task_eval/44_microwave_bowl_loading/None/summary.json",
739
+ "summarySha256": "7a3ffbc3719269037947d8916a484d07feb99706c18566308895c02dfe3e24f3"
740
+ }
741
+ },
742
+ "Equi.": {
743
+ "status": "reported",
744
+ "episodes": 50,
745
+ "successes": 49,
746
+ "lscr": 99.0,
747
+ "source": {
748
+ "runId": "medex15-official-scene-20261118-44-cov_only",
749
+ "perEpisodeFile": "per_task_eval/44_microwave_bowl_loading/Equi/per_episode.jsonl",
750
+ "perEpisodeSha256": "3e8041397b95cdfb7cf8a637dd1c58c6c7b72373089269b1df53dd7eb3bf3032",
751
+ "summaryFile": "per_task_eval/44_microwave_bowl_loading/Equi/summary.json",
752
+ "summarySha256": "71e4d66628f74e8d193cb701d9c199add0b42b6b52d9da06582e46deaa2416be"
753
+ }
754
+ },
755
+ "Inv.": {
756
+ "status": "reported",
757
+ "episodes": 50,
758
+ "successes": 43,
759
+ "lscr": 93.5,
760
+ "source": {
761
+ "runId": "medex15-official-scene-20261118-44-inv_only",
762
+ "perEpisodeFile": "per_task_eval/44_microwave_bowl_loading/Inv/per_episode.jsonl",
763
+ "perEpisodeSha256": "82678c093bf53e56428f49e06216635a2ef808e7a7e1ef4bc809cb1882cd3e1f",
764
+ "summaryFile": "per_task_eval/44_microwave_bowl_loading/Inv/summary.json",
765
+ "summarySha256": "5678fb8e3146de548ea4ace1f0365287fc9f7e230202adf721b57fd7d108ab8f"
766
+ }
767
+ },
768
+ "Full": {
769
+ "status": "reported",
770
+ "episodes": 50,
771
+ "successes": 42,
772
+ "lscr": 91.0,
773
+ "source": {
774
+ "runId": "medex15-official-scene-20261118-44-inv_cov",
775
+ "perEpisodeFile": "per_task_eval/44_microwave_bowl_loading/Full/per_episode.jsonl",
776
+ "perEpisodeSha256": "82d25163a4e28a4099620452953cb12d62fcb9c3e48026458b06f971b854a61e",
777
+ "summaryFile": "per_task_eval/44_microwave_bowl_loading/Full/summary.json",
778
+ "summarySha256": "405efc26a0f0db1fe9594c373a666d88a8a1388c3dece127dae28b7b5936d643"
779
+ }
780
+ }
781
+ },
782
+ "mean": {
783
+ "status": "reported",
784
+ "sr": 90.5,
785
+ "srScope": "four-channel-mean",
786
+ "srEpisodes": 200,
787
+ "lscr": 95.0,
788
+ "lscrScope": "four-channel-mean",
789
+ "lscrEpisodes": 200,
790
+ "efficiency": 1.0187878482436292,
791
+ "efficiencyStatus": "reported"
792
+ }
793
+ },
794
+ "51": {
795
+ "channels": {
796
+ "None": {
797
+ "status": "reported",
798
+ "episodes": 50,
799
+ "successes": 41,
800
+ "lscr": 86.66666666666667,
801
+ "source": {
802
+ "runId": "medex15-official-scene-20261118-51-none",
803
+ "perEpisodeFile": "per_task_eval/51_toilet_lid_cleaner_pour/None/per_episode.jsonl",
804
+ "perEpisodeSha256": "3ef37f536aeb3c3e71477172ac93fba92b4b5c5740d3bd3cad1826bde44201b0",
805
+ "summaryFile": "per_task_eval/51_toilet_lid_cleaner_pour/None/summary.json",
806
+ "summarySha256": "c2a974e92630d71c4820f4239d5550d7497df6f5b48a0a2ae1672e9c608ce386"
807
+ }
808
+ },
809
+ "Equi.": {
810
+ "status": "reported",
811
+ "episodes": 50,
812
+ "successes": 42,
813
+ "lscr": 88.0,
814
+ "source": {
815
+ "runId": "medex15-official-scene-20261118-51-cov_only",
816
+ "perEpisodeFile": "per_task_eval/51_toilet_lid_cleaner_pour/Equi/per_episode.jsonl",
817
+ "perEpisodeSha256": "bbe16b151553b61b9eb655fdf54bb2bb069da527d852aa360bb420866b3f0a79",
818
+ "summaryFile": "per_task_eval/51_toilet_lid_cleaner_pour/Equi/summary.json",
819
+ "summarySha256": "6de5f134cae5218432e24a65f7bc8fa38d391b2bf7d41d6d6f5c575b5a491558"
820
+ }
821
+ },
822
+ "Inv.": {
823
+ "status": "reported",
824
+ "episodes": 50,
825
+ "successes": 36,
826
+ "lscr": 79.33333333333333,
827
+ "source": {
828
+ "runId": "medex15-official-scene-20261118-51-inv_only",
829
+ "perEpisodeFile": "per_task_eval/51_toilet_lid_cleaner_pour/Inv/per_episode.jsonl",
830
+ "perEpisodeSha256": "2effef9c39a195405963a6b8150ee3eec49fe1b5e8137da1c008dbf956833b16",
831
+ "summaryFile": "per_task_eval/51_toilet_lid_cleaner_pour/Inv/summary.json",
832
+ "summarySha256": "0653ab2cc2b90927b8e71c1672577a828c81329c440fb79de05061f21678e974"
833
+ }
834
+ },
835
+ "Full": {
836
+ "status": "reported",
837
+ "episodes": 50,
838
+ "successes": 39,
839
+ "lscr": 82.0,
840
+ "source": {
841
+ "runId": "medex15-official-scene-20261118-51-inv_cov",
842
+ "perEpisodeFile": "per_task_eval/51_toilet_lid_cleaner_pour/Full/per_episode.jsonl",
843
+ "perEpisodeSha256": "e31aaf29fb8c4315d87b3c04ca30d3436854110406b6c44e1649adae96e2db9d",
844
+ "summaryFile": "per_task_eval/51_toilet_lid_cleaner_pour/Full/summary.json",
845
+ "summarySha256": "6bd4decccd86c5bd6fd4196c4f6a90db078e48a9015721306b528a3dd1ed86af"
846
+ }
847
+ }
848
+ },
849
+ "mean": {
850
+ "status": "reported",
851
+ "sr": 79,
852
+ "srScope": "four-channel-mean",
853
+ "srEpisodes": 200,
854
+ "lscr": 84.0,
855
+ "lscrScope": "four-channel-mean",
856
+ "lscrEpisodes": 200,
857
+ "efficiency": 1.1066880289212062,
858
+ "efficiencyStatus": "reported"
859
+ }
860
+ },
861
+ "60": {
862
+ "channels": {
863
+ "None": {
864
+ "status": "reported",
865
+ "episodes": 50,
866
+ "successes": 35,
867
+ "lscr": 93.5,
868
+ "source": {
869
+ "runId": "medex15-official-scene-20261118-60-none",
870
+ "perEpisodeFile": "per_task_eval/60_breadbasket_fast_food_loading/None/per_episode.jsonl",
871
+ "perEpisodeSha256": "0606ad0817bf2958e7826d45ec6f92f591f3d6f2058306c2df211c3e31f55bff",
872
+ "summaryFile": "per_task_eval/60_breadbasket_fast_food_loading/None/summary.json",
873
+ "summarySha256": "efdda1170e1a772895de6073c223c53f3d24bc6a7188031cef8b6cc717154a60"
874
+ }
875
+ },
876
+ "Equi.": {
877
+ "status": "reported",
878
+ "episodes": 50,
879
+ "successes": 32,
880
+ "lscr": 89.5,
881
+ "source": {
882
+ "runId": "medex15-official-scene-20261118-60-cov_only",
883
+ "perEpisodeFile": "per_task_eval/60_breadbasket_fast_food_loading/Equi/per_episode.jsonl",
884
+ "perEpisodeSha256": "30010bd5394db5820bb8a83fe9e2d6cc3b06a8f7b37e94f5028e959058d578a4",
885
+ "summaryFile": "per_task_eval/60_breadbasket_fast_food_loading/Equi/summary.json",
886
+ "summarySha256": "aaea0504b13b2991b3f65541dc50352bfe900b93449007e48d7e2681f6a6a7ed"
887
+ }
888
+ },
889
+ "Inv.": {
890
+ "status": "reported",
891
+ "episodes": 50,
892
+ "successes": 34,
893
+ "lscr": 85.5,
894
+ "source": {
895
+ "runId": "medex15-official-scene-20261118-60-inv_only",
896
+ "perEpisodeFile": "per_task_eval/60_breadbasket_fast_food_loading/Inv/per_episode.jsonl",
897
+ "perEpisodeSha256": "a0d8fde9e1cdc3a1022a68fc9df73187c6c2c55c4cab9dc77a2f3449a184bd20",
898
+ "summaryFile": "per_task_eval/60_breadbasket_fast_food_loading/Inv/summary.json",
899
+ "summarySha256": "408857cbfa50eab4211f88ce7c6a47a574278604b143950f9cce1c183fb34050"
900
+ }
901
+ },
902
+ "Full": {
903
+ "status": "reported",
904
+ "episodes": 50,
905
+ "successes": 32,
906
+ "lscr": 86.0,
907
+ "source": {
908
+ "runId": "medex15-official-scene-20261118-60-inv_cov",
909
+ "perEpisodeFile": "per_task_eval/60_breadbasket_fast_food_loading/Full/per_episode.jsonl",
910
+ "perEpisodeSha256": "d75c7c3bf39a538992f782bf7f40bc2b9060693bc509bd500663c881912ab5af",
911
+ "summaryFile": "per_task_eval/60_breadbasket_fast_food_loading/Full/summary.json",
912
+ "summarySha256": "c1c95f9e4fa1e4841c2753ed58b1533e37376e994153c46d7f86249040926e2e"
913
+ }
914
+ }
915
+ },
916
+ "mean": {
917
+ "status": "reported",
918
+ "sr": 66.5,
919
+ "srScope": "four-channel-mean",
920
+ "srEpisodes": 200,
921
+ "lscr": 88.625,
922
+ "lscrScope": "four-channel-mean",
923
+ "lscrEpisodes": 200,
924
+ "efficiency": 0.9466209107760132,
925
+ "efficiencyStatus": "reported"
926
+ }
927
+ },
928
+ "61": {
929
+ "channels": {
930
+ "None": {
931
+ "status": "reported",
932
+ "episodes": 50,
933
+ "successes": 35,
934
+ "lscr": 81.33333333333333,
935
+ "source": {
936
+ "runId": "medex15-official-scene-20261118-61-none",
937
+ "perEpisodeFile": "per_task_eval/61_medicine_shoebox_pack/None/per_episode.jsonl",
938
+ "perEpisodeSha256": "ed23fdeb412679fa3b86d40569bb2296f1104d0c1c199879817926a8cb873a34",
939
+ "summaryFile": "per_task_eval/61_medicine_shoebox_pack/None/summary.json",
940
+ "summarySha256": "d29bd91d4482b1b35ff114286831624003c13665f1769a4490dbd404ce94d603"
941
+ }
942
+ },
943
+ "Equi.": {
944
+ "status": "reported",
945
+ "episodes": 50,
946
+ "successes": 27,
947
+ "lscr": 71.33333333333334,
948
+ "source": {
949
+ "runId": "medex15-official-scene-20261118-61-cov_only",
950
+ "perEpisodeFile": "per_task_eval/61_medicine_shoebox_pack/Equi/per_episode.jsonl",
951
+ "perEpisodeSha256": "2056bc5f8fdc6b3eb0fcff7cdfb801dc6626558e1b4a6da9be927d5f7c1b6140",
952
+ "summaryFile": "per_task_eval/61_medicine_shoebox_pack/Equi/summary.json",
953
+ "summarySha256": "251e78b88c778a6131804de9814c0c489f2fc681e77a03feeb8a53341cc91b8a"
954
+ }
955
+ },
956
+ "Inv.": {
957
+ "status": "reported",
958
+ "episodes": 50,
959
+ "successes": 33,
960
+ "lscr": 76.66666666666666,
961
+ "source": {
962
+ "runId": "medex15-official-scene-20261118-61-inv_only",
963
+ "perEpisodeFile": "per_task_eval/61_medicine_shoebox_pack/Inv/per_episode.jsonl",
964
+ "perEpisodeSha256": "3d0988fb2b485c301e82602a06cc30847d89763e3eb96bbd3f8b31a440aaf461",
965
+ "summaryFile": "per_task_eval/61_medicine_shoebox_pack/Inv/summary.json",
966
+ "summarySha256": "6f2fcf48f873d4e52bb4ff28f7b3d682cf654ec7f097a37d0d946465dbb0d996"
967
+ }
968
+ },
969
+ "Full": {
970
+ "status": "reported",
971
+ "episodes": 50,
972
+ "successes": 30,
973
+ "lscr": 73.33333333333333,
974
+ "source": {
975
+ "runId": "medex15-official-scene-20261118-61-inv_cov",
976
+ "perEpisodeFile": "per_task_eval/61_medicine_shoebox_pack/Full/per_episode.jsonl",
977
+ "perEpisodeSha256": "bb4e3be9cd4d3d703e6a1db303bc861c8ec77b295705c2db41ea17ef3e8a404b",
978
+ "summaryFile": "per_task_eval/61_medicine_shoebox_pack/Full/summary.json",
979
+ "summarySha256": "dbc74bc7c8a2b76ec7b1d76b4a47ef5243fda2281b5cdd6a5b0efcb2d3f5155e"
980
+ }
981
+ }
982
+ },
983
+ "mean": {
984
+ "status": "reported",
985
+ "sr": 62.5,
986
+ "srScope": "four-channel-mean",
987
+ "srEpisodes": 200,
988
+ "lscr": 75.66666666666666,
989
+ "lscrScope": "four-channel-mean",
990
+ "lscrEpisodes": 200,
991
+ "efficiency": 1.0519854890129352,
992
+ "efficiencyStatus": "reported"
993
+ }
994
+ },
995
+ "62": {
996
+ "channels": {
997
+ "None": {
998
+ "status": "reported",
999
+ "episodes": 50,
1000
+ "successes": 42,
1001
+ "lscr": 90.66666666666666,
1002
+ "source": {
1003
+ "runId": "medex15-official-scene-20261118-62-none",
1004
+ "perEpisodeFile": "per_task_eval/62_shoebox_accessory_pack/None/per_episode.jsonl",
1005
+ "perEpisodeSha256": "2e31143dea46a448f08b63f8f66c30e54261eb5abaf39ce80fddc90c8560e54e",
1006
+ "summaryFile": "per_task_eval/62_shoebox_accessory_pack/None/summary.json",
1007
+ "summarySha256": "474fbcf4e72974d0f06f14653a8ff95d8196ab47d847049418318ea49d9cb3f0"
1008
+ }
1009
+ },
1010
+ "Equi.": {
1011
+ "status": "reported",
1012
+ "episodes": 50,
1013
+ "successes": 36,
1014
+ "lscr": 78.0,
1015
+ "source": {
1016
+ "runId": "medex15-official-scene-20261118-62-cov_only",
1017
+ "perEpisodeFile": "per_task_eval/62_shoebox_accessory_pack/Equi/per_episode.jsonl",
1018
+ "perEpisodeSha256": "f6f475f348c6b8d9976204f16270a20bf9209b60a2a85459409aa6f7af230857",
1019
+ "summaryFile": "per_task_eval/62_shoebox_accessory_pack/Equi/summary.json",
1020
+ "summarySha256": "b287820efe583db8d57f467540e86e4eb1d98fe79cc3a4e198faa3cddece70f6"
1021
+ }
1022
+ },
1023
+ "Inv.": {
1024
+ "status": "reported",
1025
+ "episodes": 50,
1026
+ "successes": 39,
1027
+ "lscr": 83.33333333333334,
1028
+ "source": {
1029
+ "runId": "medex15-official-scene-20261118-62-inv_only",
1030
+ "perEpisodeFile": "per_task_eval/62_shoebox_accessory_pack/Inv/per_episode.jsonl",
1031
+ "perEpisodeSha256": "02ceea2f5b7eaf74f1ce41d09b7ffd10ed94533385ef636cfa8b8d8e4cb3ade9",
1032
+ "summaryFile": "per_task_eval/62_shoebox_accessory_pack/Inv/summary.json",
1033
+ "summarySha256": "208ba3a4e1ae74541673b52ad2809b0dc6d0e14d4f21bd96c3cc69e56a984e96"
1034
+ }
1035
+ },
1036
+ "Full": {
1037
+ "status": "reported",
1038
+ "episodes": 50,
1039
+ "successes": 37,
1040
+ "lscr": 80.66666666666666,
1041
+ "source": {
1042
+ "runId": "medex15-official-scene-20261118-62-inv_cov",
1043
+ "perEpisodeFile": "per_task_eval/62_shoebox_accessory_pack/Full/per_episode.jsonl",
1044
+ "perEpisodeSha256": "337cec933c0082aff078077eccbfac0cb0960d3abe2377589c1c6d0cee7a2d3d",
1045
+ "summaryFile": "per_task_eval/62_shoebox_accessory_pack/Full/summary.json",
1046
+ "summarySha256": "4098700e118e15e45e9da82a43f6f8b090baf7e91985b8029c6f3ca26c3cc708"
1047
+ }
1048
+ }
1049
+ },
1050
+ "mean": {
1051
+ "status": "reported",
1052
+ "sr": 77,
1053
+ "srScope": "four-channel-mean",
1054
+ "srEpisodes": 200,
1055
+ "lscr": 83.16666666666666,
1056
+ "lscrScope": "four-channel-mean",
1057
+ "lscrEpisodes": 200,
1058
+ "efficiency": 0.9705257614324585,
1059
+ "efficiencyStatus": "reported"
1060
+ }
1061
+ },
1062
+ "64": {
1063
+ "channels": {
1064
+ "None": {
1065
+ "status": "reported",
1066
+ "episodes": 50,
1067
+ "successes": 36,
1068
+ "lscr": 88.5,
1069
+ "source": {
1070
+ "runId": "medex15-official-scene-20261118-64-none",
1071
+ "perEpisodeFile": "per_task_eval/64_sports_ball_cup_sort/None/per_episode.jsonl",
1072
+ "perEpisodeSha256": "26096b771a42c57c630417112232befd1b4a963bbc9fcc1538af632e660a21db",
1073
+ "summaryFile": "per_task_eval/64_sports_ball_cup_sort/None/summary.json",
1074
+ "summarySha256": "6352cf8c407d8e7ab144a2ea5d1e53f411ef3a71d2489788621338d8385f2c92"
1075
+ }
1076
+ },
1077
+ "Equi.": {
1078
+ "status": "reported",
1079
+ "episodes": 50,
1080
+ "successes": 16,
1081
+ "lscr": 59.5,
1082
+ "source": {
1083
+ "runId": "medex15-official-scene-20261118-64-cov_only",
1084
+ "perEpisodeFile": "per_task_eval/64_sports_ball_cup_sort/Equi/per_episode.jsonl",
1085
+ "perEpisodeSha256": "c808b1205201ece83366702b81c904461111ad91dcbd5524f64ab60e081ac685",
1086
+ "summaryFile": "per_task_eval/64_sports_ball_cup_sort/Equi/summary.json",
1087
+ "summarySha256": "3e1034d5ac16e891d24494119405ddc1b4e5ff4259f4a67b20f83703c863eb7b"
1088
+ }
1089
+ },
1090
+ "Inv.": {
1091
+ "status": "reported",
1092
+ "episodes": 50,
1093
+ "successes": 12,
1094
+ "lscr": 65.5,
1095
+ "source": {
1096
+ "runId": "medex15-official-scene-20261118-64-inv_only",
1097
+ "perEpisodeFile": "per_task_eval/64_sports_ball_cup_sort/Inv/per_episode.jsonl",
1098
+ "perEpisodeSha256": "cb377e5c9903b83192ad23f77d3d8df36f39680b9458599b9dd770b2b88faf82",
1099
+ "summaryFile": "per_task_eval/64_sports_ball_cup_sort/Inv/summary.json",
1100
+ "summarySha256": "b9b38278585f5d99ae324eb795c31a6873ad7a37c5796303e238ed457a37cb53"
1101
+ }
1102
+ },
1103
+ "Full": {
1104
+ "status": "reported",
1105
+ "episodes": 50,
1106
+ "successes": 11,
1107
+ "lscr": 57.99999999999999,
1108
+ "source": {
1109
+ "runId": "medex15-official-scene-20261118-64-inv_cov",
1110
+ "perEpisodeFile": "per_task_eval/64_sports_ball_cup_sort/Full/per_episode.jsonl",
1111
+ "perEpisodeSha256": "07ffa7cf7ece69335fd35833a9f92db3b2db10107df689826553979e225fb2c2",
1112
+ "summaryFile": "per_task_eval/64_sports_ball_cup_sort/Full/summary.json",
1113
+ "summarySha256": "df0aaed3bf748c4cd811dc3716b8853e6488c17557a05296080f1484b9c5f8b5"
1114
+ }
1115
+ }
1116
+ },
1117
+ "mean": {
1118
+ "status": "reported",
1119
+ "sr": 37.5,
1120
+ "srScope": "four-channel-mean",
1121
+ "srEpisodes": 200,
1122
+ "lscr": 67.875,
1123
+ "lscrScope": "four-channel-mean",
1124
+ "lscrEpisodes": 200,
1125
+ "efficiency": 0.7726265037827111,
1126
+ "efficiencyStatus": "reported"
1127
+ }
1128
+ },
1129
+ "67": {
1130
+ "channels": {
1131
+ "None": {
1132
+ "status": "reported",
1133
+ "episodes": 50,
1134
+ "successes": 37,
1135
+ "lscr": 84.5,
1136
+ "source": {
1137
+ "runId": "medex15-official-scene-20261118-67-none",
1138
+ "perEpisodeFile": "per_task_eval/67_faucet_cup_water_fill/None/per_episode.jsonl",
1139
+ "perEpisodeSha256": "799a7154c64e83a51738f22b0d8f47d6cce94a0f50002bddf4a01c87b6ffb4ed",
1140
+ "summaryFile": "per_task_eval/67_faucet_cup_water_fill/None/summary.json",
1141
+ "summarySha256": "19ba5dae940d1c4a61f1c4b2f1a4d46902ad889704ee6456a7ec68a4191c32ac"
1142
+ }
1143
+ },
1144
+ "Equi.": {
1145
+ "status": "reported",
1146
+ "episodes": 50,
1147
+ "successes": 36,
1148
+ "lscr": 85.5,
1149
+ "source": {
1150
+ "runId": "medex15-official-scene-20261118-67-cov_only",
1151
+ "perEpisodeFile": "per_task_eval/67_faucet_cup_water_fill/Equi/per_episode.jsonl",
1152
+ "perEpisodeSha256": "e63e3f8c595333e2f5fb7b37f0c7d00af234103aec56ff4f244be8b1d245c92a",
1153
+ "summaryFile": "per_task_eval/67_faucet_cup_water_fill/Equi/summary.json",
1154
+ "summarySha256": "e93162da019157d874d1a0a82fce777e644442b015e07810e8e221774699e4b9"
1155
+ }
1156
+ },
1157
+ "Inv.": {
1158
+ "status": "reported",
1159
+ "episodes": 50,
1160
+ "successes": 37,
1161
+ "lscr": 81.0,
1162
+ "source": {
1163
+ "runId": "medex15-official-scene-20261118-67-inv_only",
1164
+ "perEpisodeFile": "per_task_eval/67_faucet_cup_water_fill/Inv/per_episode.jsonl",
1165
+ "perEpisodeSha256": "917a32b463c87672f97dbae517177106cb727b283182056c23c7103168e1b6e2",
1166
+ "summaryFile": "per_task_eval/67_faucet_cup_water_fill/Inv/summary.json",
1167
+ "summarySha256": "3af47091eda5552dfbcbc24f3ba5d58e4f0210adb25aaf00e8b28cba0e1a697c"
1168
+ }
1169
+ },
1170
+ "Full": {
1171
+ "status": "reported",
1172
+ "episodes": 50,
1173
+ "successes": 36,
1174
+ "lscr": 83.5,
1175
+ "source": {
1176
+ "runId": "medex15-official-scene-20261118-67-inv_cov",
1177
+ "perEpisodeFile": "per_task_eval/67_faucet_cup_water_fill/Full/per_episode.jsonl",
1178
+ "perEpisodeSha256": "92b476bfe7ba468940e38c0647fb4bf33510e4aca4fca6a029e5f1eecb5c049d",
1179
+ "summaryFile": "per_task_eval/67_faucet_cup_water_fill/Full/summary.json",
1180
+ "summarySha256": "007e2430b62ebb573fd610877a36dd090d30cf8152bcd9f87dd8fa5a289fffe3"
1181
+ }
1182
+ }
1183
+ },
1184
+ "mean": {
1185
+ "status": "reported",
1186
+ "sr": 73,
1187
+ "srScope": "four-channel-mean",
1188
+ "srEpisodes": 200,
1189
+ "lscr": 83.625,
1190
+ "lscrScope": "four-channel-mean",
1191
+ "lscrEpisodes": 200,
1192
+ "efficiency": 0.8926039143063712,
1193
+ "efficiencyStatus": "reported"
1194
+ }
1195
+ },
1196
+ "73": {
1197
+ "channels": {
1198
+ "None": {
1199
+ "status": "reported",
1200
+ "episodes": 50,
1201
+ "successes": 0,
1202
+ "lscr": 58.5,
1203
+ "source": {
1204
+ "runId": "medex15-official-scene-20261118-73-none",
1205
+ "perEpisodeFile": "per_task_eval/73_jigsaw_puzzle_assembly/None/per_episode.jsonl",
1206
+ "perEpisodeSha256": "c6046e081e1edd464adbac00b98c3280175aea0de8a7533d55145e1c77db1c5d",
1207
+ "summaryFile": "per_task_eval/73_jigsaw_puzzle_assembly/None/summary.json",
1208
+ "summarySha256": "fe47ba18ae645d3757324d9454fc7d4d5e20f1b36c0d6ff82fe467c73688c8c2"
1209
+ }
1210
+ },
1211
+ "Equi.": {
1212
+ "status": "reported",
1213
+ "episodes": 50,
1214
+ "successes": 0,
1215
+ "lscr": 42.5,
1216
+ "source": {
1217
+ "runId": "medex15-official-scene-20261118-73-cov_only",
1218
+ "perEpisodeFile": "per_task_eval/73_jigsaw_puzzle_assembly/Equi/per_episode.jsonl",
1219
+ "perEpisodeSha256": "b12505716d8cf14270f0e3769d41b01d8c296d5e54e07a2441183b494bdcd968",
1220
+ "summaryFile": "per_task_eval/73_jigsaw_puzzle_assembly/Equi/summary.json",
1221
+ "summarySha256": "c40bea6bf9a6aab0575b6a933bf0486f5061bade8e1a63715496e702d10a664f"
1222
+ }
1223
+ },
1224
+ "Inv.": {
1225
+ "status": "reported",
1226
+ "episodes": 50,
1227
+ "successes": 0,
1228
+ "lscr": 56.99999999999999,
1229
+ "source": {
1230
+ "runId": "medex15-official-scene-20261118-73-inv_only",
1231
+ "perEpisodeFile": "per_task_eval/73_jigsaw_puzzle_assembly/Inv/per_episode.jsonl",
1232
+ "perEpisodeSha256": "dcdc98b573bd2a6b79a4f65a6bab3721df66baedbe9050da8b0e54e9b3420f47",
1233
+ "summaryFile": "per_task_eval/73_jigsaw_puzzle_assembly/Inv/summary.json",
1234
+ "summarySha256": "b30d2c430062af4e620473465884de08fe90f727024efe9346202c61901e309b"
1235
+ }
1236
+ },
1237
+ "Full": {
1238
+ "status": "reported",
1239
+ "episodes": 50,
1240
+ "successes": 1,
1241
+ "lscr": 42.0,
1242
+ "source": {
1243
+ "runId": "medex15-official-scene-20261118-73-inv_cov",
1244
+ "perEpisodeFile": "per_task_eval/73_jigsaw_puzzle_assembly/Full/per_episode.jsonl",
1245
+ "perEpisodeSha256": "1a660c52a7a3cef2e19434fc8ae786cc5c3f438e593c1f706cf5eaac5b42f1b5",
1246
+ "summaryFile": "per_task_eval/73_jigsaw_puzzle_assembly/Full/summary.json",
1247
+ "summarySha256": "132a66a2726bbafb9419b0c48beefc1ecf05c6eeff94034d10567f9d1b8c8e7d"
1248
+ }
1249
+ }
1250
+ },
1251
+ "mean": {
1252
+ "status": "reported",
1253
+ "sr": 0.5,
1254
+ "srScope": "four-channel-mean",
1255
+ "srEpisodes": 200,
1256
+ "lscr": 50.0,
1257
+ "lscrScope": "four-channel-mean",
1258
+ "lscrEpisodes": 200,
1259
+ "efficiency": 1.114889705882353,
1260
+ "efficiencyStatus": "reported"
1261
+ }
1262
+ },
1263
+ "76": {
1264
+ "channels": {
1265
+ "None": {
1266
+ "status": "reported",
1267
+ "episodes": 50,
1268
+ "successes": 9,
1269
+ "lscr": 45.0,
1270
+ "source": {
1271
+ "runId": "medex15-official-scene-20261118-76-none",
1272
+ "perEpisodeFile": "per_task_eval/76_soup_serving/None/per_episode.jsonl",
1273
+ "perEpisodeSha256": "f3315cb007065bd76bc2dc364f7f47b3ce61c9beb8a26640e3bef60f0156e51c",
1274
+ "summaryFile": "per_task_eval/76_soup_serving/None/summary.json",
1275
+ "summarySha256": "494f50579edb5c3e699ba97489f85625a4dbad478eef808f61162f39e4711e94"
1276
+ }
1277
+ },
1278
+ "Equi.": {
1279
+ "status": "reported",
1280
+ "episodes": 50,
1281
+ "successes": 7,
1282
+ "lscr": 38.5,
1283
+ "source": {
1284
+ "runId": "medex15-official-scene-20261118-76-cov_only",
1285
+ "perEpisodeFile": "per_task_eval/76_soup_serving/Equi/per_episode.jsonl",
1286
+ "perEpisodeSha256": "0a116d0670c8e7939826b4b051f4b7028ef02c5bfef24c20d5e7420ea399cb08",
1287
+ "summaryFile": "per_task_eval/76_soup_serving/Equi/summary.json",
1288
+ "summarySha256": "a2d31101f6c25bffbe71c21ea68e271ee7c95c0251b7eedc932d3bab35d10e26"
1289
+ }
1290
+ },
1291
+ "Inv.": {
1292
+ "status": "reported",
1293
+ "episodes": 50,
1294
+ "successes": 7,
1295
+ "lscr": 44.5,
1296
+ "source": {
1297
+ "runId": "medex15-official-scene-20261118-76-inv_only",
1298
+ "perEpisodeFile": "per_task_eval/76_soup_serving/Inv/per_episode.jsonl",
1299
+ "perEpisodeSha256": "34eedd0fbfd9cd17148bbaf51cfaea39383952137f893596c97d1f21ddf6a270",
1300
+ "summaryFile": "per_task_eval/76_soup_serving/Inv/summary.json",
1301
+ "summarySha256": "e1aea70561bade18dafe0b4f67d36aea7c729517219af664bcc786892cfae0b2"
1302
+ }
1303
+ },
1304
+ "Full": {
1305
+ "status": "reported",
1306
+ "episodes": 50,
1307
+ "successes": 2,
1308
+ "lscr": 38.0,
1309
+ "source": {
1310
+ "runId": "medex15-official-scene-20261118-76-inv_cov",
1311
+ "perEpisodeFile": "per_task_eval/76_soup_serving/Full/per_episode.jsonl",
1312
+ "perEpisodeSha256": "e91b0aa8ab7290d55f526082f23533602a9d3f6d4da15cebe9398ff790654a1d",
1313
+ "summaryFile": "per_task_eval/76_soup_serving/Full/summary.json",
1314
+ "summarySha256": "2f88cd308ae8cee78c0f67ac916cb81ef1b44f15fd0783d8f14f41d2436dc973"
1315
+ }
1316
+ }
1317
+ },
1318
+ "mean": {
1319
+ "status": "reported",
1320
+ "sr": 12.5,
1321
+ "srScope": "four-channel-mean",
1322
+ "srEpisodes": 200,
1323
+ "lscr": 41.5,
1324
+ "lscrScope": "four-channel-mean",
1325
+ "lscrEpisodes": 200,
1326
+ "efficiency": 0.9734810339039947,
1327
+ "efficiencyStatus": "reported"
1328
+ }
1329
+ },
1330
+ "79": {
1331
+ "channels": {
1332
+ "None": {
1333
+ "status": "reported",
1334
+ "episodes": 50,
1335
+ "successes": 14,
1336
+ "lscr": 35.0,
1337
+ "source": {
1338
+ "runId": "medex15-official-scene-20261118-79-none",
1339
+ "perEpisodeFile": "per_task_eval/79_bimanual_piano_melody/None/per_episode.jsonl",
1340
+ "perEpisodeSha256": "0642a16790714467ce38dee49b32f2ab74ce679a4ab5efd36e70bea624f7a3d0",
1341
+ "summaryFile": "per_task_eval/79_bimanual_piano_melody/None/summary.json",
1342
+ "summarySha256": "bf100795b41ebedde2921b741aa7b326dd44f7b1068b0d03e93fb86e19815a76"
1343
+ }
1344
+ },
1345
+ "Equi.": {
1346
+ "status": "reported",
1347
+ "episodes": 50,
1348
+ "successes": 17,
1349
+ "lscr": 36.0,
1350
+ "source": {
1351
+ "runId": "medex15-official-scene-20261118-79-cov_only",
1352
+ "perEpisodeFile": "per_task_eval/79_bimanual_piano_melody/Equi/per_episode.jsonl",
1353
+ "perEpisodeSha256": "0620accdb134d8b6cb84bda77d4519f7cee07d296daa3416a3bffc4fd3b7f605",
1354
+ "summaryFile": "per_task_eval/79_bimanual_piano_melody/Equi/summary.json",
1355
+ "summarySha256": "af8016b1263eeadb4e7645a11e21197ba5bb0dec83238a72cb73714267d6b27f"
1356
+ }
1357
+ },
1358
+ "Inv.": {
1359
+ "status": "reported",
1360
+ "episodes": 50,
1361
+ "successes": 14,
1362
+ "lscr": 28.999999999999996,
1363
+ "source": {
1364
+ "runId": "medex15-official-scene-20261118-79-inv_only",
1365
+ "perEpisodeFile": "per_task_eval/79_bimanual_piano_melody/Inv/per_episode.jsonl",
1366
+ "perEpisodeSha256": "0940c0488b6f07c61151b16c2a1d76d091add838490f9047c5602e5c011565b6",
1367
+ "summaryFile": "per_task_eval/79_bimanual_piano_melody/Inv/summary.json",
1368
+ "summarySha256": "b1228b6d673b2081b602a3335ff104259732ff44b17b98962405c9b17b28d64e"
1369
+ }
1370
+ },
1371
+ "Full": {
1372
+ "status": "reported",
1373
+ "episodes": 50,
1374
+ "successes": 19,
1375
+ "lscr": 41.0,
1376
+ "source": {
1377
+ "runId": "medex15-official-scene-20261118-79-inv_cov",
1378
+ "perEpisodeFile": "per_task_eval/79_bimanual_piano_melody/Full/per_episode.jsonl",
1379
+ "perEpisodeSha256": "65a5f671f8eaf2b6b4765654b62f9dd10bdf6090cc530a8444ff78a1ab674c75",
1380
+ "summaryFile": "per_task_eval/79_bimanual_piano_melody/Full/summary.json",
1381
+ "summarySha256": "cac8991b8821ddc1750fab0ab18210a0625a7478e3b6988f34392a61ea195873"
1382
+ }
1383
+ }
1384
+ },
1385
+ "mean": {
1386
+ "status": "reported",
1387
+ "sr": 32,
1388
+ "srScope": "four-channel-mean",
1389
+ "srEpisodes": 200,
1390
+ "lscr": 35.25,
1391
+ "lscrScope": "four-channel-mean",
1392
+ "lscrEpisodes": 200,
1393
+ "efficiency": 0.9377037188437913,
1394
+ "efficiencyStatus": "reported"
1395
+ }
1396
+ },
1397
+ "80": {
1398
+ "channels": {
1399
+ "None": {
1400
+ "status": "reported",
1401
+ "episodes": 50,
1402
+ "successes": 36,
1403
+ "lscr": 74.66666666666667,
1404
+ "source": {
1405
+ "runId": "medex15-official-scene-20261118-80-none",
1406
+ "perEpisodeFile": "per_task_eval/80_gaming_desk_setup/None/per_episode.jsonl",
1407
+ "perEpisodeSha256": "e991f5417b7a1a25dc7bbe70b53497ef6b5f57fe6bca1168cb4da0059480f058",
1408
+ "summaryFile": "per_task_eval/80_gaming_desk_setup/None/summary.json",
1409
+ "summarySha256": "bcc60923dc3a25ec1f497f89aabc5519fa3e265a6e9d74a7c127187eac3b29c3"
1410
+ }
1411
+ },
1412
+ "Equi.": {
1413
+ "status": "reported",
1414
+ "episodes": 50,
1415
+ "successes": 34,
1416
+ "lscr": 73.33333333333333,
1417
+ "source": {
1418
+ "runId": "medex15-official-scene-20261118-80-cov_only",
1419
+ "perEpisodeFile": "per_task_eval/80_gaming_desk_setup/Equi/per_episode.jsonl",
1420
+ "perEpisodeSha256": "f9aaa314dc628ef4fbed620261d4b1f09bf2e213c83db88c9e68359aa98dd6df",
1421
+ "summaryFile": "per_task_eval/80_gaming_desk_setup/Equi/summary.json",
1422
+ "summarySha256": "3677acb3df004c2e484f816b8f76b6efba3b29ded9811f38f48e332272ceddba"
1423
+ }
1424
+ },
1425
+ "Inv.": {
1426
+ "status": "reported",
1427
+ "episodes": 50,
1428
+ "successes": 33,
1429
+ "lscr": 71.33333333333334,
1430
+ "source": {
1431
+ "runId": "medex15-official-scene-20261118-80-inv_only",
1432
+ "perEpisodeFile": "per_task_eval/80_gaming_desk_setup/Inv/per_episode.jsonl",
1433
+ "perEpisodeSha256": "f2d1233d30148e51d23b8f0833389bada07d30f236179eb5733710c73b456211",
1434
+ "summaryFile": "per_task_eval/80_gaming_desk_setup/Inv/summary.json",
1435
+ "summarySha256": "2970d90d411a7045e7d9e12d16a3b5a92705f96561ef2a4e2e604a58991e6b8b"
1436
+ }
1437
+ },
1438
+ "Full": {
1439
+ "status": "reported",
1440
+ "episodes": 50,
1441
+ "successes": 30,
1442
+ "lscr": 68.0,
1443
+ "source": {
1444
+ "runId": "medex15-official-scene-20261118-80-inv_cov",
1445
+ "perEpisodeFile": "per_task_eval/80_gaming_desk_setup/Full/per_episode.jsonl",
1446
+ "perEpisodeSha256": "cfa1caf4f079bbe2b5968a5d9011998e0052274a297ca8bb9fcd83013a817a51",
1447
+ "summaryFile": "per_task_eval/80_gaming_desk_setup/Full/summary.json",
1448
+ "summarySha256": "326c83b40aecc02382a4a59d886f803f98230a331bc4350ff91c6b6e30fc90ee"
1449
+ }
1450
+ }
1451
+ },
1452
+ "mean": {
1453
+ "status": "reported",
1454
+ "sr": 66.5,
1455
+ "srScope": "four-channel-mean",
1456
+ "srEpisodes": 200,
1457
+ "lscr": 71.83333333333334,
1458
+ "lscrScope": "four-channel-mean",
1459
+ "lscrEpisodes": 200,
1460
+ "efficiency": 1.047898617695361,
1461
+ "efficiencyStatus": "reported"
1462
+ }
1463
+ },
1464
+ "03": {
1465
+ "channels": {
1466
+ "None": {
1467
+ "status": "reported",
1468
+ "episodes": 50,
1469
+ "successes": 25,
1470
+ "lscr": 88.66666666666666,
1471
+ "source": {
1472
+ "runId": "medex15-official-scene-20261118-03-none",
1473
+ "perEpisodeFile": "per_task_eval/03_wine_glass_plate_balance/None/per_episode.jsonl",
1474
+ "perEpisodeSha256": "33a7110be3d06a81a9b127fb0aac430977cca6526859ea17f24a81be81ffc67d",
1475
+ "summaryFile": "per_task_eval/03_wine_glass_plate_balance/None/summary.json",
1476
+ "summarySha256": "7170e7c40009e33a53cae47e4ee370ffbf632781036dc478299b8b84081e51bd"
1477
+ }
1478
+ },
1479
+ "Equi.": {
1480
+ "status": "reported",
1481
+ "episodes": 50,
1482
+ "successes": 21,
1483
+ "lscr": 87.33333333333333,
1484
+ "source": {
1485
+ "runId": "medex15-official-scene-20261118-03-cov_only",
1486
+ "perEpisodeFile": "per_task_eval/03_wine_glass_plate_balance/Equi/per_episode.jsonl",
1487
+ "perEpisodeSha256": "7033d5814181bf90dd69d0f869aba48cc4a98e19877d061a05978772394939ec",
1488
+ "summaryFile": "per_task_eval/03_wine_glass_plate_balance/Equi/summary.json",
1489
+ "summarySha256": "39624a8789b98eaa7790d256f956831ad0eb6e340ca21b5f3c0cd416a82e614c"
1490
+ }
1491
+ },
1492
+ "Inv.": {
1493
+ "status": "reported",
1494
+ "episodes": 50,
1495
+ "successes": 18,
1496
+ "lscr": 86.0,
1497
+ "source": {
1498
+ "runId": "medex15-official-scene-20261118-03-inv_only",
1499
+ "perEpisodeFile": "per_task_eval/03_wine_glass_plate_balance/Inv/per_episode.jsonl",
1500
+ "perEpisodeSha256": "6b4b39a8db429e336961f002b276b35586c7f4beb74d97aa9fe726f2f74ddd9e",
1501
+ "summaryFile": "per_task_eval/03_wine_glass_plate_balance/Inv/summary.json",
1502
+ "summarySha256": "590a53205d186cb1fb4a9ed196fa9e23b7cbac69fc44e213cfcf95dad96d88f8"
1503
+ }
1504
+ },
1505
+ "Full": {
1506
+ "status": "reported",
1507
+ "episodes": 50,
1508
+ "successes": 26,
1509
+ "lscr": 89.33333333333333,
1510
+ "source": {
1511
+ "runId": "medex15-official-scene-20261118-03-inv_cov",
1512
+ "perEpisodeFile": "per_task_eval/03_wine_glass_plate_balance/Full/per_episode.jsonl",
1513
+ "perEpisodeSha256": "3d5594e6b76ef17b39aaff5eefb8c84ec90b1a780199d4808b0a586b978d643b",
1514
+ "summaryFile": "per_task_eval/03_wine_glass_plate_balance/Full/summary.json",
1515
+ "summarySha256": "e8cb1d34851e0ddd13ef70ba4b4505ed8e4a7f3a416d3fc6b7200a6ce594f301"
1516
+ }
1517
+ }
1518
+ },
1519
+ "mean": {
1520
+ "status": "reported",
1521
+ "sr": 45,
1522
+ "srScope": "four-channel-mean",
1523
+ "srEpisodes": 200,
1524
+ "lscr": 87.83333333333333,
1525
+ "lscrScope": "four-channel-mean",
1526
+ "lscrEpisodes": 200,
1527
+ "efficiency": 1.0480828736714651,
1528
+ "efficiencyStatus": "reported"
1529
+ }
1530
+ },
1531
+ "06": {
1532
+ "channels": {
1533
+ "None": {
1534
+ "status": "reported",
1535
+ "episodes": 50,
1536
+ "successes": 49,
1537
+ "lscr": 100.0,
1538
+ "source": {
1539
+ "runId": "medex15-official-scene-20261118-06-none",
1540
+ "perEpisodeFile": "per_task_eval/06_fruit_bowl_loading/None/per_episode.jsonl",
1541
+ "perEpisodeSha256": "052505e35b9f6c2a8f877c18e4ce46d7aced29785701fd085a1579f2c3d594a0",
1542
+ "summaryFile": "per_task_eval/06_fruit_bowl_loading/None/summary.json",
1543
+ "summarySha256": "7b3e28566d6dd9de0450690e83ff102de2a780084495eacd9a65ee0e9dd7af03"
1544
+ }
1545
+ },
1546
+ "Equi.": {
1547
+ "status": "reported",
1548
+ "episodes": 50,
1549
+ "successes": 39,
1550
+ "lscr": 96.0,
1551
+ "source": {
1552
+ "runId": "medex15-official-scene-20261118-06-cov_only",
1553
+ "perEpisodeFile": "per_task_eval/06_fruit_bowl_loading/Equi/per_episode.jsonl",
1554
+ "perEpisodeSha256": "1450ca73dafe8fcaf81fce3336609152eb0f24d5a46f016cf4cbab72c2ccfb80",
1555
+ "summaryFile": "per_task_eval/06_fruit_bowl_loading/Equi/summary.json",
1556
+ "summarySha256": "83a287cc367cfa84e8f4bf8b8a37efb00803469a5268d21e58eeb20a7cd0c097"
1557
+ }
1558
+ },
1559
+ "Inv.": {
1560
+ "status": "reported",
1561
+ "episodes": 50,
1562
+ "successes": 41,
1563
+ "lscr": 97.5,
1564
+ "source": {
1565
+ "runId": "medex15-official-scene-20261118-06-inv_only",
1566
+ "perEpisodeFile": "per_task_eval/06_fruit_bowl_loading/Inv/per_episode.jsonl",
1567
+ "perEpisodeSha256": "eeeb82a4bfb36eb95a9d73940c8fa66eb346e89c06f19ea9dec5d11876f7d8ef",
1568
+ "summaryFile": "per_task_eval/06_fruit_bowl_loading/Inv/summary.json",
1569
+ "summarySha256": "e98429f99643980b481bd8a3ea345b7ee4e823c2f1d85b12ce1e60c514cc28a4"
1570
+ }
1571
+ },
1572
+ "Full": {
1573
+ "status": "reported",
1574
+ "episodes": 50,
1575
+ "successes": 45,
1576
+ "lscr": 98.0,
1577
+ "source": {
1578
+ "runId": "medex15-official-scene-20261118-06-inv_cov",
1579
+ "perEpisodeFile": "per_task_eval/06_fruit_bowl_loading/Full/per_episode.jsonl",
1580
+ "perEpisodeSha256": "05b9d18ffa5343cacc10315cc9d0239912913f9ba3974e792a618417b5db7b1b",
1581
+ "summaryFile": "per_task_eval/06_fruit_bowl_loading/Full/summary.json",
1582
+ "summarySha256": "463f2881771df3038680797e402d1a532536f00d5b63415c60cc3f4839a10829"
1583
+ }
1584
+ }
1585
+ },
1586
+ "mean": {
1587
+ "status": "reported",
1588
+ "sr": 87,
1589
+ "srScope": "four-channel-mean",
1590
+ "srEpisodes": 200,
1591
+ "lscr": 97.875,
1592
+ "lscrScope": "four-channel-mean",
1593
+ "lscrEpisodes": 200,
1594
+ "efficiency": 0.9300577527780214,
1595
+ "efficiencyStatus": "reported"
1596
+ }
1597
+ },
1598
+ "07": {
1599
+ "channels": {
1600
+ "None": {
1601
+ "status": "reported",
1602
+ "episodes": 50,
1603
+ "successes": 46,
1604
+ "lscr": 98.5,
1605
+ "source": {
1606
+ "runId": "medex15-official-scene-20261118-07-none",
1607
+ "perEpisodeFile": "per_task_eval/07_citrus_plate_loading/None/per_episode.jsonl",
1608
+ "perEpisodeSha256": "9779abf2d409bb90478b5ed4fb8c72d2c5c5feb38f93bf3b09929353f80f4043",
1609
+ "summaryFile": "per_task_eval/07_citrus_plate_loading/None/summary.json",
1610
+ "summarySha256": "126c9f0054f18f73a4b9772b69f3ae01a5089e0fca8ba8f4f8715a78748c9786"
1611
+ }
1612
+ },
1613
+ "Equi.": {
1614
+ "status": "reported",
1615
+ "episodes": 50,
1616
+ "successes": 46,
1617
+ "lscr": 97.0,
1618
+ "source": {
1619
+ "runId": "medex15-official-scene-20261118-07-cov_only",
1620
+ "perEpisodeFile": "per_task_eval/07_citrus_plate_loading/Equi/per_episode.jsonl",
1621
+ "perEpisodeSha256": "0afff7b34b7f983d95cfb774d57d09ce8c1b79c07a354e0eeb42814bf61b804b",
1622
+ "summaryFile": "per_task_eval/07_citrus_plate_loading/Equi/summary.json",
1623
+ "summarySha256": "b6fa3cecf2b98dde83235c55caf6bb9a67f8221375feac8b54618587ff186240"
1624
+ }
1625
+ },
1626
+ "Inv.": {
1627
+ "status": "reported",
1628
+ "episodes": 50,
1629
+ "successes": 44,
1630
+ "lscr": 96.0,
1631
+ "source": {
1632
+ "runId": "medex15-official-scene-20261118-07-inv_only",
1633
+ "perEpisodeFile": "per_task_eval/07_citrus_plate_loading/Inv/per_episode.jsonl",
1634
+ "perEpisodeSha256": "17652b7c77cd61054603a0ddbb62abd7ed0a268dedf2a779c9324d5f5b69e8e0",
1635
+ "summaryFile": "per_task_eval/07_citrus_plate_loading/Inv/summary.json",
1636
+ "summarySha256": "cd0d9b2b693be5873e00c30dd76c89efe7cedb26a9a7d06745008e81f4036110"
1637
+ }
1638
+ },
1639
+ "Full": {
1640
+ "status": "reported",
1641
+ "episodes": 50,
1642
+ "successes": 47,
1643
+ "lscr": 98.5,
1644
+ "source": {
1645
+ "runId": "medex15-official-scene-20261118-07-inv_cov",
1646
+ "perEpisodeFile": "per_task_eval/07_citrus_plate_loading/Full/per_episode.jsonl",
1647
+ "perEpisodeSha256": "4fd3274652ce8a441cd7b1f645a81e4c3e8e486d8b54e5b884ebcc104b3f5ad8",
1648
+ "summaryFile": "per_task_eval/07_citrus_plate_loading/Full/summary.json",
1649
+ "summarySha256": "0dab0041de2ea4c394edcd4799d6be33087e4432cf439512486d6c4df555b18b"
1650
+ }
1651
+ }
1652
+ },
1653
+ "mean": {
1654
+ "status": "reported",
1655
+ "sr": 91.5,
1656
+ "srScope": "four-channel-mean",
1657
+ "srEpisodes": 200,
1658
+ "lscr": 97.5,
1659
+ "lscrScope": "four-channel-mean",
1660
+ "lscrEpisodes": 200,
1661
+ "efficiency": 0.8370779894827698,
1662
+ "efficiencyStatus": "reported"
1663
+ }
1664
+ },
1665
+ "08": {
1666
+ "channels": {
1667
+ "None": {
1668
+ "status": "reported",
1669
+ "episodes": 50,
1670
+ "successes": 48,
1671
+ "lscr": 97.0,
1672
+ "source": {
1673
+ "runId": "medex15-official-scene-20261118-08-none",
1674
+ "perEpisodeFile": "per_task_eval/08_frypan_stand_pour/None/per_episode.jsonl",
1675
+ "perEpisodeSha256": "d5012a7cc83892d75cd6231063fe8fd3ffd57c8927ec54ca4283568e3a2e79dd",
1676
+ "summaryFile": "per_task_eval/08_frypan_stand_pour/None/summary.json",
1677
+ "summarySha256": "e6977117f1f5961cda6c1069d1a63b5c41f07d12cbdd2ffbf4c1e21e108db9fa"
1678
+ }
1679
+ },
1680
+ "Equi.": {
1681
+ "status": "reported",
1682
+ "episodes": 50,
1683
+ "successes": 46,
1684
+ "lscr": 96.0,
1685
+ "source": {
1686
+ "runId": "medex15-official-scene-20261118-08-cov_only",
1687
+ "perEpisodeFile": "per_task_eval/08_frypan_stand_pour/Equi/per_episode.jsonl",
1688
+ "perEpisodeSha256": "074990a5bdccff1d290bed4f6b32d11c1c89106827a492d5731a19403a9feebd",
1689
+ "summaryFile": "per_task_eval/08_frypan_stand_pour/Equi/summary.json",
1690
+ "summarySha256": "173e1075fb41c296aa7c645558d667f576224ab4e513ff4cb00688250fa9c092"
1691
+ }
1692
+ },
1693
+ "Inv.": {
1694
+ "status": "reported",
1695
+ "episodes": 50,
1696
+ "successes": 48,
1697
+ "lscr": 97.0,
1698
+ "source": {
1699
+ "runId": "medex15-official-scene-20261118-08-inv_only",
1700
+ "perEpisodeFile": "per_task_eval/08_frypan_stand_pour/Inv/per_episode.jsonl",
1701
+ "perEpisodeSha256": "72b749e35e520706c2185918d664c450f741a906ee09ca389051afe449a93c4b",
1702
+ "summaryFile": "per_task_eval/08_frypan_stand_pour/Inv/summary.json",
1703
+ "summarySha256": "e0b8ce4d01ab2c09c041c09801c0dd40440d5e76c27512955fa72ae854573377"
1704
+ }
1705
+ },
1706
+ "Full": {
1707
+ "status": "reported",
1708
+ "episodes": 50,
1709
+ "successes": 46,
1710
+ "lscr": 95.0,
1711
+ "source": {
1712
+ "runId": "medex15-official-scene-20261118-08-inv_cov",
1713
+ "perEpisodeFile": "per_task_eval/08_frypan_stand_pour/Full/per_episode.jsonl",
1714
+ "perEpisodeSha256": "ae0ee903d6f49a8a5c80afd755ca86f15609e594cb3e02ca7a3a7c345af7f10c",
1715
+ "summaryFile": "per_task_eval/08_frypan_stand_pour/Full/summary.json",
1716
+ "summarySha256": "636a9f382fa4a0a35c4b7d519d3aba138c900d0e53f280d0c2a464e0a37e45aa"
1717
+ }
1718
+ }
1719
+ },
1720
+ "mean": {
1721
+ "status": "reported",
1722
+ "sr": 94,
1723
+ "srScope": "four-channel-mean",
1724
+ "srEpisodes": 200,
1725
+ "lscr": 96.25,
1726
+ "lscrScope": "four-channel-mean",
1727
+ "lscrEpisodes": 200,
1728
+ "efficiency": 1.3134649208520617,
1729
+ "efficiencyStatus": "reported"
1730
+ }
1731
+ },
1732
+ "09": {
1733
+ "channels": {
1734
+ "None": {
1735
+ "status": "reported",
1736
+ "episodes": 50,
1737
+ "successes": 45,
1738
+ "lscr": 95.0,
1739
+ "source": {
1740
+ "runId": "medex15-official-scene-20261118-09-none",
1741
+ "perEpisodeFile": "per_task_eval/09_cleaner_moisturizer_box_loading/None/per_episode.jsonl",
1742
+ "perEpisodeSha256": "a8859faf448cc5a086cac513c4c0f6e8367936a2054032bf7bb2325faa557eec",
1743
+ "summaryFile": "per_task_eval/09_cleaner_moisturizer_box_loading/None/summary.json",
1744
+ "summarySha256": "59defadc7b2d385646b5526032f4426e2b1b3025557ce2f7d9cea4dad3fa8c34"
1745
+ }
1746
+ },
1747
+ "Equi.": {
1748
+ "status": "reported",
1749
+ "episodes": 50,
1750
+ "successes": 35,
1751
+ "lscr": 95.0,
1752
+ "source": {
1753
+ "runId": "medex15-official-scene-20261118-09-cov_only",
1754
+ "perEpisodeFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Equi/per_episode.jsonl",
1755
+ "perEpisodeSha256": "b119a2294f51193b68870884056b0beae83ce47a81e7ab40cfd5c6f1eef773a7",
1756
+ "summaryFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Equi/summary.json",
1757
+ "summarySha256": "6954d4b2cf1b592924c689574a482709c68257a7f59ef969c2bd3ecc4f3be35f"
1758
+ }
1759
+ },
1760
+ "Inv.": {
1761
+ "status": "reported",
1762
+ "episodes": 50,
1763
+ "successes": 34,
1764
+ "lscr": 93.0,
1765
+ "source": {
1766
+ "runId": "medex15-official-scene-20261118-09-inv_only",
1767
+ "perEpisodeFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Inv/per_episode.jsonl",
1768
+ "perEpisodeSha256": "30cd5e71ce920f9e5d26297c666cea407d79a2f5b4629c1f041996ff8f08dc9c",
1769
+ "summaryFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Inv/summary.json",
1770
+ "summarySha256": "f866327165df8fa27c5574eadaf477da64cd194ff8d4680db4c5a20dbc4b944e"
1771
+ }
1772
+ },
1773
+ "Full": {
1774
+ "status": "reported",
1775
+ "episodes": 50,
1776
+ "successes": 31,
1777
+ "lscr": 92.0,
1778
+ "source": {
1779
+ "runId": "medex15-official-scene-20261118-09-inv_cov",
1780
+ "perEpisodeFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Full/per_episode.jsonl",
1781
+ "perEpisodeSha256": "4ba0af2e4b34fbb4ba6ea6b54dcfc2c21a148582222cc1bf26c8ae270796950d",
1782
+ "summaryFile": "per_task_eval/09_cleaner_moisturizer_box_loading/Full/summary.json",
1783
+ "summarySha256": "9254606bc1830892fd8a37b4f9cb497b8f204e99fdb416b4f86fd6a8c73668dd"
1784
+ }
1785
+ }
1786
+ },
1787
+ "mean": {
1788
+ "status": "reported",
1789
+ "sr": 72.5,
1790
+ "srScope": "four-channel-mean",
1791
+ "srEpisodes": 200,
1792
+ "lscr": 93.75,
1793
+ "lscrScope": "four-channel-mean",
1794
+ "lscrEpisodes": 200,
1795
+ "efficiency": 0.9092353107924499,
1796
+ "efficiencyStatus": "reported"
1797
+ }
1798
+ }
1799
+ },
1800
+ "evaluation": {
1801
+ "schemaVersion": "1.0",
1802
+ "isTemplate": false,
1803
+ "protocol": {
1804
+ "benchmarkCommit": "c269f24bd39bd973e157a16c3c9c666cf5f43efc",
1805
+ "taskCount": 26,
1806
+ "channels": [
1807
+ "None",
1808
+ "Equi.",
1809
+ "Inv.",
1810
+ "Full"
1811
+ ],
1812
+ "episodesPerChannel": 50,
1813
+ "seeds": {
1814
+ "03": {
1815
+ "None": {
1816
+ "base": 100300000,
1817
+ "first": 1,
1818
+ "last": 50
1819
+ },
1820
+ "Equi.": {
1821
+ "base": 100300000,
1822
+ "first": 1,
1823
+ "last": 50
1824
+ },
1825
+ "Inv.": {
1826
+ "base": 100300000,
1827
+ "first": 1,
1828
+ "last": 50
1829
+ },
1830
+ "Full": {
1831
+ "base": 200300000,
1832
+ "first": 1,
1833
+ "last": 50
1834
+ }
1835
+ },
1836
+ "06": {
1837
+ "None": {
1838
+ "base": 100600000,
1839
+ "first": 1,
1840
+ "last": 50
1841
+ },
1842
+ "Equi.": {
1843
+ "base": 100600000,
1844
+ "first": 1,
1845
+ "last": 50
1846
+ },
1847
+ "Inv.": {
1848
+ "base": 100600000,
1849
+ "first": 1,
1850
+ "last": 50
1851
+ },
1852
+ "Full": {
1853
+ "base": 200600000,
1854
+ "first": 1,
1855
+ "last": 50
1856
+ }
1857
+ },
1858
+ "07": {
1859
+ "None": {
1860
+ "base": 100700000,
1861
+ "first": 1,
1862
+ "last": 50
1863
+ },
1864
+ "Equi.": {
1865
+ "base": 100700000,
1866
+ "first": 1,
1867
+ "last": 50
1868
+ },
1869
+ "Inv.": {
1870
+ "base": 100700000,
1871
+ "first": 1,
1872
+ "last": 50
1873
+ },
1874
+ "Full": {
1875
+ "base": 200700000,
1876
+ "first": 1,
1877
+ "last": 50
1878
+ }
1879
+ },
1880
+ "08": {
1881
+ "None": {
1882
+ "base": 100800000,
1883
+ "first": 1,
1884
+ "last": 50
1885
+ },
1886
+ "Equi.": {
1887
+ "base": 100800000,
1888
+ "first": 1,
1889
+ "last": 50
1890
+ },
1891
+ "Inv.": {
1892
+ "base": 100800000,
1893
+ "first": 1,
1894
+ "last": 50
1895
+ },
1896
+ "Full": {
1897
+ "base": 200800000,
1898
+ "first": 1,
1899
+ "last": 50
1900
+ }
1901
+ },
1902
+ "09": {
1903
+ "None": {
1904
+ "base": 100900000,
1905
+ "first": 1,
1906
+ "last": 50
1907
+ },
1908
+ "Equi.": {
1909
+ "base": 100900000,
1910
+ "first": 1,
1911
+ "last": 50
1912
+ },
1913
+ "Inv.": {
1914
+ "base": 100900000,
1915
+ "first": 1,
1916
+ "last": 50
1917
+ },
1918
+ "Full": {
1919
+ "base": 200900000,
1920
+ "first": 1,
1921
+ "last": 50
1922
+ }
1923
+ },
1924
+ "12": {
1925
+ "None": {
1926
+ "base": 101200000,
1927
+ "first": 1,
1928
+ "last": 50
1929
+ },
1930
+ "Equi.": {
1931
+ "base": 101200000,
1932
+ "first": 1,
1933
+ "last": 50
1934
+ },
1935
+ "Inv.": {
1936
+ "base": 101200000,
1937
+ "first": 1,
1938
+ "last": 50
1939
+ },
1940
+ "Full": {
1941
+ "base": 201200000,
1942
+ "first": 1,
1943
+ "last": 50
1944
+ }
1945
+ },
1946
+ "21": {
1947
+ "None": {
1948
+ "base": 102100000,
1949
+ "first": 1,
1950
+ "last": 50
1951
+ },
1952
+ "Equi.": {
1953
+ "base": 102100000,
1954
+ "first": 1,
1955
+ "last": 50
1956
+ },
1957
+ "Inv.": {
1958
+ "base": 102100000,
1959
+ "first": 1,
1960
+ "last": 50
1961
+ },
1962
+ "Full": {
1963
+ "base": 202100000,
1964
+ "first": 1,
1965
+ "last": 50
1966
+ }
1967
+ },
1968
+ "22": {
1969
+ "None": {
1970
+ "base": 102200000,
1971
+ "first": 1,
1972
+ "last": 50
1973
+ },
1974
+ "Equi.": {
1975
+ "base": 102200000,
1976
+ "first": 1,
1977
+ "last": 50
1978
+ },
1979
+ "Inv.": {
1980
+ "base": 102200000,
1981
+ "first": 1,
1982
+ "last": 50
1983
+ },
1984
+ "Full": {
1985
+ "base": 202200000,
1986
+ "first": 1,
1987
+ "last": 50
1988
+ }
1989
+ },
1990
+ "24": {
1991
+ "None": {
1992
+ "base": 102400000,
1993
+ "first": 1,
1994
+ "last": 50
1995
+ },
1996
+ "Equi.": {
1997
+ "base": 102400000,
1998
+ "first": 1,
1999
+ "last": 50
2000
+ },
2001
+ "Inv.": {
2002
+ "base": 102400000,
2003
+ "first": 1,
2004
+ "last": 50
2005
+ },
2006
+ "Full": {
2007
+ "base": 202400000,
2008
+ "first": 1,
2009
+ "last": 50
2010
+ }
2011
+ },
2012
+ "26": {
2013
+ "None": {
2014
+ "base": 102600000,
2015
+ "first": 1,
2016
+ "last": 50
2017
+ },
2018
+ "Equi.": {
2019
+ "base": 102600000,
2020
+ "first": 1,
2021
+ "last": 50
2022
+ },
2023
+ "Inv.": {
2024
+ "base": 102600000,
2025
+ "first": 1,
2026
+ "last": 50
2027
+ },
2028
+ "Full": {
2029
+ "base": 202600000,
2030
+ "first": 1,
2031
+ "last": 50
2032
+ }
2033
+ },
2034
+ "27": {
2035
+ "None": {
2036
+ "base": 102700000,
2037
+ "first": 1,
2038
+ "last": 50
2039
+ },
2040
+ "Equi.": {
2041
+ "base": 102700000,
2042
+ "first": 1,
2043
+ "last": 50
2044
+ },
2045
+ "Inv.": {
2046
+ "base": 102700000,
2047
+ "first": 1,
2048
+ "last": 50
2049
+ },
2050
+ "Full": {
2051
+ "base": 202700000,
2052
+ "first": 1,
2053
+ "last": 50
2054
+ }
2055
+ },
2056
+ "32": {
2057
+ "None": {
2058
+ "base": 103200000,
2059
+ "first": 1,
2060
+ "last": 50
2061
+ },
2062
+ "Equi.": {
2063
+ "base": 103200000,
2064
+ "first": 1,
2065
+ "last": 50
2066
+ },
2067
+ "Inv.": {
2068
+ "base": 103200000,
2069
+ "first": 1,
2070
+ "last": 50
2071
+ },
2072
+ "Full": {
2073
+ "base": 203200000,
2074
+ "first": 1,
2075
+ "last": 50
2076
+ }
2077
+ },
2078
+ "34": {
2079
+ "None": {
2080
+ "base": 103400000,
2081
+ "first": 1,
2082
+ "last": 50
2083
+ },
2084
+ "Equi.": {
2085
+ "base": 103400000,
2086
+ "first": 1,
2087
+ "last": 50
2088
+ },
2089
+ "Inv.": {
2090
+ "base": 103400000,
2091
+ "first": 1,
2092
+ "last": 50
2093
+ },
2094
+ "Full": {
2095
+ "base": 203400000,
2096
+ "first": 1,
2097
+ "last": 50
2098
+ }
2099
+ },
2100
+ "42": {
2101
+ "None": {
2102
+ "base": 104200000,
2103
+ "first": 1,
2104
+ "last": 50
2105
+ },
2106
+ "Equi.": {
2107
+ "base": 104200000,
2108
+ "first": 1,
2109
+ "last": 50
2110
+ },
2111
+ "Inv.": {
2112
+ "base": 104200000,
2113
+ "first": 1,
2114
+ "last": 50
2115
+ },
2116
+ "Full": {
2117
+ "base": 204200000,
2118
+ "first": 1,
2119
+ "last": 50
2120
+ }
2121
+ },
2122
+ "43": {
2123
+ "None": {
2124
+ "base": 104300000,
2125
+ "first": 1,
2126
+ "last": 50
2127
+ },
2128
+ "Equi.": {
2129
+ "base": 104300000,
2130
+ "first": 1,
2131
+ "last": 50
2132
+ },
2133
+ "Inv.": {
2134
+ "base": 104300000,
2135
+ "first": 1,
2136
+ "last": 50
2137
+ },
2138
+ "Full": {
2139
+ "base": 204300000,
2140
+ "first": 1,
2141
+ "last": 50
2142
+ }
2143
+ },
2144
+ "44": {
2145
+ "None": {
2146
+ "base": 104400000,
2147
+ "first": 1,
2148
+ "last": 50
2149
+ },
2150
+ "Equi.": {
2151
+ "base": 104400000,
2152
+ "first": 1,
2153
+ "last": 50
2154
+ },
2155
+ "Inv.": {
2156
+ "base": 104400000,
2157
+ "first": 1,
2158
+ "last": 50
2159
+ },
2160
+ "Full": {
2161
+ "base": 204400000,
2162
+ "first": 1,
2163
+ "last": 50
2164
+ }
2165
+ },
2166
+ "51": {
2167
+ "None": {
2168
+ "base": 105100000,
2169
+ "first": 1,
2170
+ "last": 50
2171
+ },
2172
+ "Equi.": {
2173
+ "base": 105100000,
2174
+ "first": 1,
2175
+ "last": 50
2176
+ },
2177
+ "Inv.": {
2178
+ "base": 105100000,
2179
+ "first": 1,
2180
+ "last": 50
2181
+ },
2182
+ "Full": {
2183
+ "base": 205100000,
2184
+ "first": 1,
2185
+ "last": 50
2186
+ }
2187
+ },
2188
+ "60": {
2189
+ "None": {
2190
+ "base": 106000000,
2191
+ "first": 1,
2192
+ "last": 50
2193
+ },
2194
+ "Equi.": {
2195
+ "base": 106000000,
2196
+ "first": 1,
2197
+ "last": 50
2198
+ },
2199
+ "Inv.": {
2200
+ "base": 106000000,
2201
+ "first": 1,
2202
+ "last": 50
2203
+ },
2204
+ "Full": {
2205
+ "base": 206000000,
2206
+ "first": 1,
2207
+ "last": 50
2208
+ }
2209
+ },
2210
+ "61": {
2211
+ "None": {
2212
+ "base": 106100000,
2213
+ "first": 1,
2214
+ "last": 50
2215
+ },
2216
+ "Equi.": {
2217
+ "base": 106100000,
2218
+ "first": 1,
2219
+ "last": 50
2220
+ },
2221
+ "Inv.": {
2222
+ "base": 106100000,
2223
+ "first": 1,
2224
+ "last": 50
2225
+ },
2226
+ "Full": {
2227
+ "base": 206100000,
2228
+ "first": 1,
2229
+ "last": 50
2230
+ }
2231
+ },
2232
+ "62": {
2233
+ "None": {
2234
+ "base": 106200000,
2235
+ "first": 1,
2236
+ "last": 50
2237
+ },
2238
+ "Equi.": {
2239
+ "base": 106200000,
2240
+ "first": 1,
2241
+ "last": 50
2242
+ },
2243
+ "Inv.": {
2244
+ "base": 106200000,
2245
+ "first": 1,
2246
+ "last": 50
2247
+ },
2248
+ "Full": {
2249
+ "base": 206200000,
2250
+ "first": 1,
2251
+ "last": 50
2252
+ }
2253
+ },
2254
+ "64": {
2255
+ "None": {
2256
+ "base": 106400000,
2257
+ "first": 1,
2258
+ "last": 50
2259
+ },
2260
+ "Equi.": {
2261
+ "base": 106400000,
2262
+ "first": 1,
2263
+ "last": 50
2264
+ },
2265
+ "Inv.": {
2266
+ "base": 106400000,
2267
+ "first": 1,
2268
+ "last": 50
2269
+ },
2270
+ "Full": {
2271
+ "base": 206400000,
2272
+ "first": 1,
2273
+ "last": 50
2274
+ }
2275
+ },
2276
+ "67": {
2277
+ "None": {
2278
+ "base": 106700000,
2279
+ "first": 1,
2280
+ "last": 50
2281
+ },
2282
+ "Equi.": {
2283
+ "base": 106700000,
2284
+ "first": 1,
2285
+ "last": 50
2286
+ },
2287
+ "Inv.": {
2288
+ "base": 106700000,
2289
+ "first": 1,
2290
+ "last": 50
2291
+ },
2292
+ "Full": {
2293
+ "base": 206700000,
2294
+ "first": 1,
2295
+ "last": 50
2296
+ }
2297
+ },
2298
+ "73": {
2299
+ "None": {
2300
+ "base": 107300000,
2301
+ "first": 1,
2302
+ "last": 50
2303
+ },
2304
+ "Equi.": {
2305
+ "base": 107300000,
2306
+ "first": 1,
2307
+ "last": 50
2308
+ },
2309
+ "Inv.": {
2310
+ "base": 107300000,
2311
+ "first": 1,
2312
+ "last": 50
2313
+ },
2314
+ "Full": {
2315
+ "base": 207300000,
2316
+ "first": 1,
2317
+ "last": 50
2318
+ }
2319
+ },
2320
+ "76": {
2321
+ "None": {
2322
+ "base": 107600000,
2323
+ "first": 1,
2324
+ "last": 50
2325
+ },
2326
+ "Equi.": {
2327
+ "base": 107600000,
2328
+ "first": 1,
2329
+ "last": 50
2330
+ },
2331
+ "Inv.": {
2332
+ "base": 107600000,
2333
+ "first": 1,
2334
+ "last": 50
2335
+ },
2336
+ "Full": {
2337
+ "base": 207600000,
2338
+ "first": 1,
2339
+ "last": 50
2340
+ }
2341
+ },
2342
+ "79": {
2343
+ "None": {
2344
+ "base": 107900000,
2345
+ "first": 1,
2346
+ "last": 50
2347
+ },
2348
+ "Equi.": {
2349
+ "base": 107900000,
2350
+ "first": 1,
2351
+ "last": 50
2352
+ },
2353
+ "Inv.": {
2354
+ "base": 107900000,
2355
+ "first": 1,
2356
+ "last": 50
2357
+ },
2358
+ "Full": {
2359
+ "base": 207900000,
2360
+ "first": 1,
2361
+ "last": 50
2362
+ }
2363
+ },
2364
+ "80": {
2365
+ "None": {
2366
+ "base": 108000000,
2367
+ "first": 1,
2368
+ "last": 50
2369
+ },
2370
+ "Equi.": {
2371
+ "base": 108000000,
2372
+ "first": 1,
2373
+ "last": 50
2374
+ },
2375
+ "Inv.": {
2376
+ "base": 108000000,
2377
+ "first": 1,
2378
+ "last": 50
2379
+ },
2380
+ "Full": {
2381
+ "base": 208000000,
2382
+ "first": 1,
2383
+ "last": 50
2384
+ }
2385
+ }
2386
+ },
2387
+ "stepUnit": "physics_steps",
2388
+ "physicsHz": 60,
2389
+ "controlHz": 20,
2390
+ "executionStepBudget": {
2391
+ "03": 3246,
2392
+ "06": 1920,
2393
+ "07": 1806,
2394
+ "08": 3432,
2395
+ "09": 1926,
2396
+ "12": 3447,
2397
+ "21": 2481,
2398
+ "22": 2043,
2399
+ "24": 4626,
2400
+ "26": 2613,
2401
+ "27": 3570,
2402
+ "32": 2556,
2403
+ "34": 3240,
2404
+ "42": 2004,
2405
+ "43": 4125,
2406
+ "44": 3183,
2407
+ "51": 3069,
2408
+ "60": 2985,
2409
+ "61": 3102,
2410
+ "62": 3231,
2411
+ "64": 1995,
2412
+ "67": 2892,
2413
+ "73": 3639,
2414
+ "76": 2958,
2415
+ "79": 2043,
2416
+ "80": 1779
2417
+ },
2418
+ "expertReference": "Official scene expert_time_step, in physics steps.",
2419
+ "protocolChanges": "Official reach_and_stop evaluation with native scene instructions, execution budgets and dwell times. One task-specialized checkpoint per task. Physical tactile force is set to zero at evaluation while the tactile encoder and expert remain active.",
2420
+ "metricSchemaVersion": "core-v1.8-policy-step-query-split"
2421
+ },
2422
+ "coverage": {
2423
+ "evaluatedTaskCount": 26,
2424
+ "evaluatedEpisodeCount": 5200,
2425
+ "completedTaskCount": {
2426
+ "None": 26,
2427
+ "Equi.": 26,
2428
+ "Inv.": 26,
2429
+ "Full": 26,
2430
+ "mean": 26
2431
+ },
2432
+ "expectedFullEpisodeCount": 1300,
2433
+ "expectedFourChannelEpisodeCount": 5200,
2434
+ "missingTaskChannels": {}
2435
+ },
2436
+ "aggregation": {
2437
+ "units": {
2438
+ "sr": "percent",
2439
+ "lscr": "percent",
2440
+ "efficiency": "ratio",
2441
+ "steps": "physics_steps"
2442
+ },
2443
+ "sr": "100 * successes / episodes per task/channel; equal channel mean per task; equal task mean over all 26 settings.",
2444
+ "lscr": "100 * mean episode latched_stage_completion_rate, including failures; equal channel mean per task; equal task mean over all 26 settings.",
2445
+ "efficiency": "expert_time_step / mean successful steps_to_stable_success. Pool successful episodes across all four channels before calculating task mean efficiency. Overall values average task efficiencies over tasks with success; report valid-task counts.",
2446
+ "missing": "null means unreported (NR). Efficiency without successes is null with no-success status. Valid-task counts are null when unreported, and 0 for an evaluated scope without successes.",
2447
+ "scope": "Overall channel aggregates require all 26 settings. Four-channel means require all four channels. Never substitute Full-only scores for means or include duplicate reruns."
2448
+ },
2449
+ "logs": {
2450
+ "repository": "Bench2Dex/Leaderboard",
2451
+ "policyPath": "me-dex-1-5",
2452
+ "revision": null,
2453
+ "status": "not-archived"
2454
+ }
2455
+ }
2456
+ }
me-dex-1-5/logo.png ADDED

Git LFS Details

  • SHA256: 15973e6740d795609287aa3b391166a7a85a6046834576de869843168853ba50
  • Pointer size: 131 Bytes
  • Size of remote file: 198 kB
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Equi/summary.json ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100300000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 21,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3246,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.42,
50
+ "success_rate_ci": {
51
+ "mean": 0.42,
52
+ "ci_low": 0.28,
53
+ "ci_high": 0.56
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.8733333333333333
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 34.565873015872334
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.3,
64
+ "hard_violation_rate": 0.6,
65
+ "drop_rate": 0.6,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 0.0002905139918801339
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.5,
74
+ "at_end_success_rate": 0.42,
75
+ "task_family_success_rate": {
76
+ "table_setting": 0.42
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.7666666666666666,
81
+ "current_chain_depth_progress_score": 0.7666666666666666,
82
+ "latched_chain_depth_progress_score": 0.8733333333333333,
83
+ "avg_current_chain_depth": 2.3,
84
+ "avg_latched_chain_depth": 2.62
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.0461621497132725,
88
+ "expert_time_s": 36.06666666666666,
89
+ "expert_time_step": 2164,
90
+ "avg_steps_to_success": 2073.9523809523807,
91
+ "avg_policy_steps_to_success": 691.7142857142857,
92
+ "avg_policy_queries_to_success": 43.666666666666664
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0002905139918801339,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_194_wineglass_1": 4,
98
+ "drop_to_table_obj_194_wineglass_2": 16,
99
+ "drop_to_table_obj_194_wineglass_3": 20
100
+ },
101
+ "robot_constraint_violation_rate": 0.2,
102
+ "finger_joint_limit_saturation_rate": 0.0,
103
+ "joint_limit_active_step_rate": 0.0002251483437071038,
104
+ "finger_joint_limit_saturation_step_rate": 0.0,
105
+ "max_joint_limit_excess_rad": 0.13374729454517365,
106
+ "robot_constraint_diagnostics": {
107
+ "joint_limit_soft_counts": {
108
+ "multi_ring_q2": 3,
109
+ "pinky_q1": 7,
110
+ "ring_q1": 3,
111
+ "ring_q2": 18
112
+ },
113
+ "joint_limit_hard_counts": {
114
+ "middle_q1": 1,
115
+ "multi_ring_q2": 2,
116
+ "pinky_q1": 14,
117
+ "pinky_q2": 3,
118
+ "ring_q1": 2,
119
+ "ring_q2": 6
120
+ },
121
+ "joint_limit_hard_episode_events": 24,
122
+ "finger_joint_limit_saturation_counts": {},
123
+ "joint_limit_max_excess_by_joint": {
124
+ "index_q1": 0.0033336332999169827,
125
+ "index_q2": 9.812205270520735e-09,
126
+ "middle_q1": 0.008638332597911358,
127
+ "middle_q2": 0.0009072992252185941,
128
+ "multi_index_q1": 0.0006522813928313553,
129
+ "multi_index_q2": 5.70031133406701e-09,
130
+ "multi_middle_q1": 0.0005643780459649861,
131
+ "multi_middle_q2": 0.00013185925490688533,
132
+ "multi_pinky_q1": 0.0019594854675233364,
133
+ "multi_pinky_q2": 0.004764404613524675,
134
+ "multi_ring_q1": 0.0034298605751246214,
135
+ "multi_ring_q2": 0.051052067428827286,
136
+ "pinky_q1": 0.01431944314390421,
137
+ "pinky_q2": 0.006593650206923485,
138
+ "ring_q1": 0.01718299463391304,
139
+ "ring_q2": 0.13374729454517365,
140
+ "thumb_q2": 0.0025597906205803156
141
+ },
142
+ "coverage": {
143
+ "hard_limit_source": {
144
+ "available": 50,
145
+ "unavailable": 0,
146
+ "unknown": 0
147
+ },
148
+ "joint_role_manifest": {
149
+ "available": 50,
150
+ "unavailable": 0,
151
+ "unknown": 0
152
+ }
153
+ },
154
+ "hard_limit_sources": {
155
+ "runtime.joint_pos_limits": 50
156
+ },
157
+ "soft_limit_sources": {
158
+ "runtime.soft_joint_pos_limits": 50
159
+ },
160
+ "unknown_runtime_joint_names": [],
161
+ "baseline_checks": {
162
+ "post_home": {
163
+ "checked": 50,
164
+ "hard_violation_episodes": 0
165
+ },
166
+ "post_reset": {
167
+ "checked": 50,
168
+ "hard_violation_episodes": 0
169
+ }
170
+ }
171
+ }
172
+ },
173
+ "robot_motion": {
174
+ "joint_acceleration_rms": 10.908733202798013,
175
+ "joint_effort_rms": 98.36767756815033,
176
+ "joint_jerk_rms": 1084.5221732966643,
177
+ "joint_velocity_rms": 0.8470196674453964
178
+ },
179
+ "grasp": {
180
+ "kinematic_grasp_stability_index": 0.9138385117059018,
181
+ "mean_kinematic_grasp_stability": 0.6372984548125195,
182
+ "per_object_best_kinematic_grasp_stability": {
183
+ "obj_194_wineglass_1": 0.8238990248858227,
184
+ "obj_194_wineglass_2": 0.8386596001585015,
185
+ "obj_194_wineglass_3": 0.7829518886021494
186
+ },
187
+ "grasp_gsi_diagnostics": {
188
+ "mean_anti_slip_score": 0.9269145425565744,
189
+ "mean_duration_score": 0.5747099759095702,
190
+ "mean_hold_score": 0.204445644578113,
191
+ "mean_kinematic_stability_score": 0.12844822889057328,
192
+ "mean_legacy_gsi": 0.41666219800607984,
193
+ "mean_lift_score": 0.22696960580655925,
194
+ "mean_motion_stability_score": 0.17895384998761898
195
+ }
196
+ },
197
+ "tool": {
198
+ "tool_selection_accuracy": null,
199
+ "tool_switch_success_rate": null,
200
+ "tool_switch_total": 0,
201
+ "tool_switch_successful": 0
202
+ },
203
+ "runtime": {
204
+ "avg_steps": 2753.74,
205
+ "error_count": 0
206
+ }
207
+ }
208
+ }
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Full/summary.json ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 200300000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 26,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3246,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.52,
50
+ "success_rate_ci": {
51
+ "mean": 0.52,
52
+ "ci_low": 0.38,
53
+ "ci_high": 0.66
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.8933333333333333
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 34.291666666666
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.4,
64
+ "hard_violation_rate": 0.56,
65
+ "drop_rate": 0.56,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 0.0002663642797890395
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.66,
74
+ "at_end_success_rate": 0.52,
75
+ "task_family_success_rate": {
76
+ "table_setting": 0.52
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.8066666666666666,
81
+ "current_chain_depth_progress_score": 0.8066666666666666,
82
+ "latched_chain_depth_progress_score": 0.8933333333333333,
83
+ "avg_current_chain_depth": 2.42,
84
+ "avg_latched_chain_depth": 2.68
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.0573448355818698,
88
+ "expert_time_s": 36.06666666666666,
89
+ "expert_time_step": 2164,
90
+ "avg_steps_to_success": 2057.5,
91
+ "avg_policy_steps_to_success": 686.2307692307693,
92
+ "avg_policy_queries_to_success": 43.46153846153846
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0002663642797890395,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_194_wineglass_1": 5,
98
+ "drop_to_table_obj_194_wineglass_2": 11,
99
+ "drop_to_table_obj_194_wineglass_3": 19
100
+ },
101
+ "robot_constraint_violation_rate": 0.08,
102
+ "finger_joint_limit_saturation_rate": 0.0,
103
+ "joint_limit_active_step_rate": 0.0,
104
+ "finger_joint_limit_saturation_step_rate": 0.0,
105
+ "max_joint_limit_excess_rad": 0.009394340217113495,
106
+ "robot_constraint_diagnostics": {
107
+ "joint_limit_soft_counts": {},
108
+ "joint_limit_hard_counts": {
109
+ "pinky_q1": 8,
110
+ "ring_q1": 1
111
+ },
112
+ "joint_limit_hard_episode_events": 9,
113
+ "finger_joint_limit_saturation_counts": {},
114
+ "joint_limit_max_excess_by_joint": {
115
+ "index_q1": 0.004109308123588562,
116
+ "index_q2": 1.1811926547977691e-08,
117
+ "middle_q1": 0.0027174949645996094,
118
+ "middle_q2": 0.000727713166270405,
119
+ "multi_index_q1": 0.0008605365292169154,
120
+ "multi_index_q2": 6.941871522059273e-09,
121
+ "multi_middle_q1": 0.0008300869376398623,
122
+ "multi_middle_q2": 0.00033664042712189257,
123
+ "multi_pinky_q1": 0.003925632685422897,
124
+ "multi_pinky_q2": 0.00031555918394587934,
125
+ "multi_ring_q1": 0.0033911506179720163,
126
+ "multi_ring_q2": 0.0007041998906061053,
127
+ "pinky_q1": 0.009394340217113495,
128
+ "pinky_q2": 0.0015374863287433982,
129
+ "ring_q1": 0.006480695679783821,
130
+ "ring_q2": 0.004686984233558178,
131
+ "thumb_q2": 0.0006980948382988572
132
+ },
133
+ "coverage": {
134
+ "hard_limit_source": {
135
+ "available": 50,
136
+ "unavailable": 0,
137
+ "unknown": 0
138
+ },
139
+ "joint_role_manifest": {
140
+ "available": 50,
141
+ "unavailable": 0,
142
+ "unknown": 0
143
+ }
144
+ },
145
+ "hard_limit_sources": {
146
+ "runtime.joint_pos_limits": 50
147
+ },
148
+ "soft_limit_sources": {
149
+ "runtime.soft_joint_pos_limits": 50
150
+ },
151
+ "unknown_runtime_joint_names": [],
152
+ "baseline_checks": {
153
+ "post_home": {
154
+ "checked": 50,
155
+ "hard_violation_episodes": 0
156
+ },
157
+ "post_reset": {
158
+ "checked": 50,
159
+ "hard_violation_episodes": 0
160
+ }
161
+ }
162
+ }
163
+ },
164
+ "robot_motion": {
165
+ "joint_acceleration_rms": 11.08355621724677,
166
+ "joint_effort_rms": 99.2338911122432,
167
+ "joint_jerk_rms": 1101.5858116379288,
168
+ "joint_velocity_rms": 0.8605311531120673
169
+ },
170
+ "grasp": {
171
+ "kinematic_grasp_stability_index": 0.8843190414630622,
172
+ "mean_kinematic_grasp_stability": 0.5909659322801639,
173
+ "per_object_best_kinematic_grasp_stability": {
174
+ "obj_194_wineglass_1": 0.8010504749447936,
175
+ "obj_194_wineglass_2": 0.8026839989270435,
176
+ "obj_194_wineglass_3": 0.7913886624524635
177
+ },
178
+ "grasp_gsi_diagnostics": {
179
+ "mean_anti_slip_score": 0.9288906429109733,
180
+ "mean_duration_score": 0.556993790889048,
181
+ "mean_hold_score": 0.17393236143608115,
182
+ "mean_kinematic_stability_score": 0.09493414638281554,
183
+ "mean_legacy_gsi": 0.39573515293386535,
184
+ "mean_lift_score": 0.19341409164983064,
185
+ "mean_motion_stability_score": 0.14956834863777196
186
+ }
187
+ },
188
+ "tool": {
189
+ "tool_selection_accuracy": null,
190
+ "tool_switch_success_rate": null,
191
+ "tool_switch_total": 0,
192
+ "tool_switch_successful": 0
193
+ },
194
+ "runtime": {
195
+ "avg_steps": 2627.98,
196
+ "error_count": 0
197
+ }
198
+ }
199
+ }
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/Inv/summary.json ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100300000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 18,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3246,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.36,
50
+ "success_rate_ci": {
51
+ "mean": 0.36,
52
+ "ci_low": 0.22,
53
+ "ci_high": 0.5
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.86
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 33.91851851851788
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.32,
64
+ "hard_violation_rate": 0.62,
65
+ "drop_rate": 0.62,
66
+ "high_speed_violation_rate": 0.02,
67
+ "safety_violation_step_rate": 0.0002989238740534077
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.5,
74
+ "at_end_success_rate": 0.36,
75
+ "task_family_success_rate": {
76
+ "table_setting": 0.36
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.74,
81
+ "current_chain_depth_progress_score": 0.74,
82
+ "latched_chain_depth_progress_score": 0.86,
83
+ "avg_current_chain_depth": 2.22,
84
+ "avg_latched_chain_depth": 2.58
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.0657653787888088,
88
+ "expert_time_s": 36.06666666666666,
89
+ "expert_time_step": 2164,
90
+ "avg_steps_to_success": 2035.111111111111,
91
+ "avg_policy_steps_to_success": 678.6666666666666,
92
+ "avg_policy_queries_to_success": 42.888888888888886
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.000291806638956898,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_194_wineglass_1": 8,
98
+ "drop_to_table_obj_194_wineglass_2": 13,
99
+ "drop_to_table_obj_194_wineglass_3": 19,
100
+ "high_speed_distractor_00_clutter_054_softball": 1
101
+ },
102
+ "robot_constraint_violation_rate": 0.14,
103
+ "finger_joint_limit_saturation_rate": 0.0,
104
+ "joint_limit_active_step_rate": 0.0003700962250185048,
105
+ "finger_joint_limit_saturation_step_rate": 0.0,
106
+ "max_joint_limit_excess_rad": 0.11578531563282013,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "ring_q1": 20,
110
+ "ring_q2": 32
111
+ },
112
+ "joint_limit_hard_counts": {
113
+ "index_q1": 5,
114
+ "pinky_q1": 7,
115
+ "ring_q1": 14,
116
+ "ring_q2": 5
117
+ },
118
+ "joint_limit_hard_episode_events": 31,
119
+ "finger_joint_limit_saturation_counts": {},
120
+ "joint_limit_max_excess_by_joint": {
121
+ "index_q1": 0.007745875045657158,
122
+ "index_q2": 9.336214468191883e-09,
123
+ "middle_q1": 0.003708974691107869,
124
+ "middle_q2": 0.003041449934244156,
125
+ "multi_index_q1": 0.00045749361743219197,
126
+ "multi_index_q2": 7.68500374448422e-09,
127
+ "multi_middle_q1": 0.0007532780291512609,
128
+ "multi_middle_q2": 8.46035618451424e-05,
129
+ "multi_pinky_q1": 0.0010935968020930886,
130
+ "multi_pinky_q2": 0.0010710732312873006,
131
+ "multi_ring_q1": 0.0039059785194694996,
132
+ "multi_ring_q2": 0.0005993424565531313,
133
+ "pinky_q1": 0.00996597483754158,
134
+ "pinky_q2": 0.0027287036646157503,
135
+ "ring_q1": 0.09253831952810287,
136
+ "ring_q2": 0.11578531563282013,
137
+ "thumb_q2": 0.000786030141171068
138
+ },
139
+ "coverage": {
140
+ "hard_limit_source": {
141
+ "available": 50,
142
+ "unavailable": 0,
143
+ "unknown": 0
144
+ },
145
+ "joint_role_manifest": {
146
+ "available": 50,
147
+ "unavailable": 0,
148
+ "unknown": 0
149
+ }
150
+ },
151
+ "hard_limit_sources": {
152
+ "runtime.joint_pos_limits": 50
153
+ },
154
+ "soft_limit_sources": {
155
+ "runtime.soft_joint_pos_limits": 50
156
+ },
157
+ "unknown_runtime_joint_names": [],
158
+ "baseline_checks": {
159
+ "post_home": {
160
+ "checked": 50,
161
+ "hard_violation_episodes": 0
162
+ },
163
+ "post_reset": {
164
+ "checked": 50,
165
+ "hard_violation_episodes": 0
166
+ }
167
+ }
168
+ }
169
+ },
170
+ "robot_motion": {
171
+ "joint_acceleration_rms": 11.265030995471728,
172
+ "joint_effort_rms": 99.06642722969147,
173
+ "joint_jerk_rms": 1119.537209368968,
174
+ "joint_velocity_rms": 0.8558953211807577
175
+ },
176
+ "grasp": {
177
+ "kinematic_grasp_stability_index": 0.9027859352451713,
178
+ "mean_kinematic_grasp_stability": 0.6375705619417646,
179
+ "per_object_best_kinematic_grasp_stability": {
180
+ "obj_194_wineglass_1": 0.8111193333587976,
181
+ "obj_194_wineglass_2": 0.8438309074023396,
182
+ "obj_194_wineglass_3": 0.7421511883814402
183
+ },
184
+ "grasp_gsi_diagnostics": {
185
+ "mean_anti_slip_score": 0.9244735431759706,
186
+ "mean_duration_score": 0.5817925560810866,
187
+ "mean_hold_score": 0.20410652489984732,
188
+ "mean_kinematic_stability_score": 0.13124497793299708,
189
+ "mean_legacy_gsi": 0.4182747335370486,
190
+ "mean_lift_score": 0.2264055337707472,
191
+ "mean_motion_stability_score": 0.1806222294052366
192
+ }
193
+ },
194
+ "tool": {
195
+ "tool_selection_accuracy": null,
196
+ "tool_switch_success_rate": null,
197
+ "tool_switch_total": 0,
198
+ "tool_switch_successful": 0
199
+ },
200
+ "runtime": {
201
+ "avg_steps": 2810.08,
202
+ "error_count": 0
203
+ }
204
+ }
205
+ }
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/03_wine_glass_plate_balance/None/summary.json ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100300000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 25,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3246,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.5,
50
+ "success_rate_ci": {
51
+ "mean": 0.5,
52
+ "ci_low": 0.36,
53
+ "ci_high": 0.64
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.8866666666666666
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 34.763333333332646
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.38,
64
+ "hard_violation_rate": 0.58,
65
+ "drop_rate": 0.58,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 0.000292584117933906
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.54,
74
+ "at_end_success_rate": 0.5,
75
+ "task_family_success_rate": {
76
+ "table_setting": 0.5
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.78,
81
+ "current_chain_depth_progress_score": 0.78,
82
+ "latched_chain_depth_progress_score": 0.8866666666666666,
83
+ "avg_current_chain_depth": 2.34,
84
+ "avg_latched_chain_depth": 2.66
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.043301959126203,
88
+ "expert_time_s": 36.06666666666666,
89
+ "expert_time_step": 2164,
90
+ "avg_steps_to_success": 2085.8,
91
+ "avg_policy_steps_to_success": 695.64,
92
+ "avg_policy_queries_to_success": 43.92
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.000292584117933906,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_194_wineglass_1": 8,
98
+ "drop_to_table_obj_194_wineglass_2": 12,
99
+ "drop_to_table_obj_194_wineglass_3": 19
100
+ },
101
+ "robot_constraint_violation_rate": 0.2,
102
+ "finger_joint_limit_saturation_rate": 0.0,
103
+ "joint_limit_active_step_rate": 0.00015004313740200306,
104
+ "finger_joint_limit_saturation_step_rate": 0.0,
105
+ "max_joint_limit_excess_rad": 0.029556626453995705,
106
+ "robot_constraint_diagnostics": {
107
+ "joint_limit_soft_counts": {
108
+ "multi_pinky_q1": 5,
109
+ "multi_ring_q1": 5,
110
+ "pinky_q1": 1,
111
+ "ring_q1": 9
112
+ },
113
+ "joint_limit_hard_counts": {
114
+ "multi_pinky_q1": 2,
115
+ "multi_ring_q1": 1,
116
+ "multi_ring_q2": 3,
117
+ "pinky_q1": 9,
118
+ "ring_q1": 6,
119
+ "ring_q2": 2
120
+ },
121
+ "joint_limit_hard_episode_events": 20,
122
+ "finger_joint_limit_saturation_counts": {},
123
+ "joint_limit_max_excess_by_joint": {
124
+ "index_q1": 0.004094071686267853,
125
+ "index_q2": 1.466846999420568e-08,
126
+ "middle_q1": 0.001962503883987665,
127
+ "middle_q2": 0.003781392704695463,
128
+ "multi_index_q1": 0.0012472550151869655,
129
+ "multi_index_q2": 6.56125509479466e-09,
130
+ "multi_middle_q1": 0.0008327449322678149,
131
+ "multi_middle_q2": 0.00011771021672757342,
132
+ "multi_pinky_q1": 0.012072333134710789,
133
+ "multi_pinky_q2": 0.0013096433831378818,
134
+ "multi_ring_q1": 0.029556626453995705,
135
+ "multi_ring_q2": 0.009677881374955177,
136
+ "pinky_q1": 0.010508393868803978,
137
+ "pinky_q2": 0.0049281916581094265,
138
+ "ring_q1": 0.02794063836336136,
139
+ "ring_q2": 0.0055412789806723595,
140
+ "thumb_q2": 0.0014355747262015939
141
+ },
142
+ "coverage": {
143
+ "hard_limit_source": {
144
+ "available": 50,
145
+ "unavailable": 0,
146
+ "unknown": 0
147
+ },
148
+ "joint_role_manifest": {
149
+ "available": 50,
150
+ "unavailable": 0,
151
+ "unknown": 0
152
+ }
153
+ },
154
+ "hard_limit_sources": {
155
+ "runtime.joint_pos_limits": 50
156
+ },
157
+ "soft_limit_sources": {
158
+ "runtime.soft_joint_pos_limits": 50
159
+ },
160
+ "unknown_runtime_joint_names": [],
161
+ "baseline_checks": {
162
+ "post_home": {
163
+ "checked": 50,
164
+ "hard_violation_episodes": 0
165
+ },
166
+ "post_reset": {
167
+ "checked": 50,
168
+ "hard_violation_episodes": 0
169
+ }
170
+ }
171
+ }
172
+ },
173
+ "robot_motion": {
174
+ "joint_acceleration_rms": 10.852282323140097,
175
+ "joint_effort_rms": 99.53472173215326,
176
+ "joint_jerk_rms": 1078.904209079041,
177
+ "joint_velocity_rms": 0.8449202108145325
178
+ },
179
+ "grasp": {
180
+ "kinematic_grasp_stability_index": 0.9001948357579336,
181
+ "mean_kinematic_grasp_stability": 0.6273295272005771,
182
+ "per_object_best_kinematic_grasp_stability": {
183
+ "obj_194_wineglass_1": 0.8098603717978813,
184
+ "obj_194_wineglass_2": 0.8467898693477709,
185
+ "obj_194_wineglass_3": 0.7672751453205053
186
+ },
187
+ "grasp_gsi_diagnostics": {
188
+ "mean_anti_slip_score": 0.9417926339525409,
189
+ "mean_duration_score": 0.55924078228487,
190
+ "mean_hold_score": 0.17743331235939122,
191
+ "mean_kinematic_stability_score": 0.117730999682303,
192
+ "mean_legacy_gsi": 0.40014878272050414,
193
+ "mean_lift_score": 0.19726988707340198,
194
+ "mean_motion_stability_score": 0.151863970504078
195
+ }
196
+ },
197
+ "tool": {
198
+ "tool_selection_accuracy": null,
199
+ "tool_switch_success_rate": null,
200
+ "tool_switch_total": 0,
201
+ "tool_switch_successful": 0
202
+ },
203
+ "runtime": {
204
+ "avg_steps": 2665.9,
205
+ "error_count": 0
206
+ }
207
+ }
208
+ }
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Equi/summary.json ADDED
@@ -0,0 +1,265 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100600000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 39,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1920,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.78,
50
+ "success_rate_ci": {
51
+ "mean": 0.78,
52
+ "ci_low": 0.66,
53
+ "ci_high": 0.88
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.96
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.306837606837565
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.48,
64
+ "hard_violation_rate": 0.52,
65
+ "drop_rate": 0.22,
66
+ "high_speed_violation_rate": 0.42,
67
+ "safety_violation_step_rate": 0.00068730339157789
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.8,
74
+ "at_end_success_rate": 0.78,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.78
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.905,
81
+ "current_chain_depth_progress_score": 0.9333333333333333,
82
+ "latched_chain_depth_progress_score": 0.96,
83
+ "avg_current_chain_depth": 2.8,
84
+ "avg_latched_chain_depth": 2.88
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9181051676779952,
88
+ "expert_time_s": 21.333333333333332,
89
+ "expert_time_step": 1280,
90
+ "avg_steps_to_success": 1398.4102564102564,
91
+ "avg_policy_steps_to_success": 466.5128205128205,
92
+ "avg_policy_queries_to_success": 29.564102564102566
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0005286949165983769,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_011_banana_2": 2,
98
+ "drop_to_table_obj_013_apple_3": 5,
99
+ "drop_to_table_obj_013_apple_4": 9,
100
+ "high_speed_obj_013_apple_3": 12,
101
+ "high_speed_obj_013_apple_4": 12
102
+ },
103
+ "robot_constraint_violation_rate": 1.0,
104
+ "finger_joint_limit_saturation_rate": 0.98,
105
+ "joint_limit_active_step_rate": 0.1580136931983399,
106
+ "finger_joint_limit_saturation_step_rate": 0.08777657352824553,
107
+ "max_joint_limit_excess_rad": 1.2847635746002197,
108
+ "robot_constraint_diagnostics": {
109
+ "joint_limit_soft_counts": {
110
+ "left_index_1_joint": 1240,
111
+ "left_index_2_joint": 33,
112
+ "left_little_1_joint": 1344,
113
+ "left_little_2_joint": 196,
114
+ "left_middle_1_joint": 1575,
115
+ "left_middle_2_joint": 111,
116
+ "left_ring_1_joint": 1625,
117
+ "left_ring_2_joint": 263,
118
+ "left_thumb_2_joint": 72,
119
+ "left_thumb_3_joint": 6,
120
+ "left_thumb_4_joint": 94,
121
+ "right_index_1_joint": 1121,
122
+ "right_index_2_joint": 7,
123
+ "right_little_1_joint": 279,
124
+ "right_little_2_joint": 94,
125
+ "right_middle_1_joint": 1490,
126
+ "right_middle_2_joint": 17,
127
+ "right_ring_1_joint": 555,
128
+ "right_ring_2_joint": 15,
129
+ "right_thumb_2_joint": 8101,
130
+ "right_thumb_3_joint": 41,
131
+ "right_thumb_4_joint": 70
132
+ },
133
+ "joint_limit_hard_counts": {
134
+ "left_index_1_joint": 66,
135
+ "left_index_2_joint": 47,
136
+ "left_little_1_joint": 53,
137
+ "left_little_2_joint": 118,
138
+ "left_middle_1_joint": 60,
139
+ "left_middle_2_joint": 140,
140
+ "left_ring_1_joint": 57,
141
+ "left_ring_2_joint": 131,
142
+ "left_thumb_2_joint": 4,
143
+ "left_thumb_3_joint": 2,
144
+ "left_thumb_4_joint": 5,
145
+ "right_index_1_joint": 59,
146
+ "right_index_2_joint": 2,
147
+ "right_little_1_joint": 21,
148
+ "right_little_2_joint": 4,
149
+ "right_middle_1_joint": 62,
150
+ "right_middle_2_joint": 8,
151
+ "right_ring_1_joint": 33,
152
+ "right_ring_2_joint": 6,
153
+ "right_thumb_2_joint": 97,
154
+ "right_thumb_3_joint": 2,
155
+ "right_thumb_4_joint": 14
156
+ },
157
+ "joint_limit_hard_episode_events": 241,
158
+ "finger_joint_limit_saturation_counts": {
159
+ "left_index_1_joint": 313,
160
+ "left_little_1_joint": 755,
161
+ "left_little_2_joint": 6,
162
+ "left_middle_1_joint": 744,
163
+ "left_ring_1_joint": 783,
164
+ "left_thumb_4_joint": 33,
165
+ "right_index_1_joint": 497,
166
+ "right_little_1_joint": 210,
167
+ "right_little_2_joint": 67,
168
+ "right_middle_1_joint": 731,
169
+ "right_ring_1_joint": 253,
170
+ "right_thumb_2_joint": 4706,
171
+ "right_thumb_3_joint": 28,
172
+ "right_thumb_4_joint": 5
173
+ },
174
+ "joint_limit_max_excess_by_joint": {
175
+ "left_index_1_joint": 0.8607226014137268,
176
+ "left_index_2_joint": 0.047572024166584015,
177
+ "left_little_1_joint": 1.272802710533142,
178
+ "left_little_2_joint": 0.24382378160953522,
179
+ "left_middle_1_joint": 0.9461638331413269,
180
+ "left_middle_2_joint": 0.044440269470214844,
181
+ "left_ring_1_joint": 1.160179853439331,
182
+ "left_ring_2_joint": 0.09843311458826065,
183
+ "left_thumb_2_joint": 0.08481799066066742,
184
+ "left_thumb_3_joint": 0.02593206986784935,
185
+ "left_thumb_4_joint": 0.5507813692092896,
186
+ "right_index_1_joint": 1.1784285306930542,
187
+ "right_index_2_joint": 0.05234205722808838,
188
+ "right_little_1_joint": 1.0581815242767334,
189
+ "right_little_2_joint": 0.7489802837371826,
190
+ "right_middle_1_joint": 1.2847635746002197,
191
+ "right_middle_2_joint": 0.06743061542510986,
192
+ "right_ring_1_joint": 1.2159829139709473,
193
+ "right_ring_2_joint": 0.0567317008972168,
194
+ "right_thumb_1_joint": 8.809566497802734e-05,
195
+ "right_thumb_2_joint": 0.7303107976913452,
196
+ "right_thumb_3_joint": 0.782504141330719,
197
+ "right_thumb_4_joint": 0.4303596317768097
198
+ },
199
+ "coverage": {
200
+ "hard_limit_source": {
201
+ "available": 50,
202
+ "unavailable": 0,
203
+ "unknown": 0
204
+ },
205
+ "joint_role_manifest": {
206
+ "available": 50,
207
+ "unavailable": 0,
208
+ "unknown": 0
209
+ }
210
+ },
211
+ "hard_limit_sources": {
212
+ "runtime.joint_pos_limits": 50
213
+ },
214
+ "soft_limit_sources": {
215
+ "runtime.soft_joint_pos_limits": 50
216
+ },
217
+ "unknown_runtime_joint_names": [],
218
+ "baseline_checks": {
219
+ "post_home": {
220
+ "checked": 50,
221
+ "hard_violation_episodes": 0
222
+ },
223
+ "post_reset": {
224
+ "checked": 50,
225
+ "hard_violation_episodes": 0
226
+ }
227
+ }
228
+ }
229
+ },
230
+ "robot_motion": {
231
+ "joint_acceleration_rms": 51.843890074074444,
232
+ "joint_effort_rms": 39.420314737716915,
233
+ "joint_jerk_rms": 5301.010201676266,
234
+ "joint_velocity_rms": 1.0695432106737366
235
+ },
236
+ "grasp": {
237
+ "kinematic_grasp_stability_index": 0.9114732753173469,
238
+ "mean_kinematic_grasp_stability": 0.3917667003132423,
239
+ "per_object_best_kinematic_grasp_stability": {
240
+ "obj_011_banana_2": 0.6847595497191539,
241
+ "obj_013_apple_3": 0.8411501870947299,
242
+ "obj_013_apple_4": 0.2493028164669721
243
+ },
244
+ "grasp_gsi_diagnostics": {
245
+ "mean_anti_slip_score": 0.9241385395288086,
246
+ "mean_duration_score": 0.13667362283291715,
247
+ "mean_hold_score": 0.13392826404866992,
248
+ "mean_kinematic_stability_score": 0.08636544179555666,
249
+ "mean_legacy_gsi": 0.29723669048590673,
250
+ "mean_lift_score": 0.22180701977561224,
251
+ "mean_motion_stability_score": 0.12600047263720976
252
+ }
253
+ },
254
+ "tool": {
255
+ "tool_selection_accuracy": null,
256
+ "tool_switch_success_rate": null,
257
+ "tool_switch_total": 0,
258
+ "tool_switch_successful": 0
259
+ },
260
+ "runtime": {
261
+ "avg_steps": 1513.16,
262
+ "error_count": 0
263
+ }
264
+ }
265
+ }
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Full/summary.json ADDED
@@ -0,0 +1,254 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 200600000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 45,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1920,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.9,
50
+ "success_rate_ci": {
51
+ "mean": 0.9,
52
+ "ci_low": 0.82,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.98
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 22.978888888888864
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.68,
64
+ "hard_violation_rate": 0.3,
65
+ "drop_rate": 0.16,
66
+ "high_speed_violation_rate": 0.3,
67
+ "safety_violation_step_rate": 0.00037686864034169426
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.9,
74
+ "at_end_success_rate": 0.9,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.9
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.965,
81
+ "current_chain_depth_progress_score": 0.9866666666666667,
82
+ "latched_chain_depth_progress_score": 0.9866666666666667,
83
+ "avg_current_chain_depth": 2.96,
84
+ "avg_latched_chain_depth": 2.96
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9317907094410834,
88
+ "expert_time_s": 21.333333333333332,
89
+ "expert_time_step": 1280,
90
+ "avg_steps_to_success": 1378.7333333333333,
91
+ "avg_policy_steps_to_success": 459.8888888888889,
92
+ "avg_policy_queries_to_success": 29.155555555555555
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0003349943469703949,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_013_apple_3": 2,
98
+ "drop_to_table_obj_013_apple_4": 6,
99
+ "high_speed_obj_013_apple_3": 8,
100
+ "high_speed_obj_013_apple_4": 8
101
+ },
102
+ "robot_constraint_violation_rate": 1.0,
103
+ "finger_joint_limit_saturation_rate": 1.0,
104
+ "joint_limit_active_step_rate": 0.16935360049132503,
105
+ "finger_joint_limit_saturation_step_rate": 0.08716832070125484,
106
+ "max_joint_limit_excess_rad": 1.4315546751022339,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "left_index_1_joint": 1238,
110
+ "left_index_2_joint": 62,
111
+ "left_little_1_joint": 1750,
112
+ "left_little_2_joint": 269,
113
+ "left_middle_1_joint": 1844,
114
+ "left_middle_2_joint": 151,
115
+ "left_ring_1_joint": 2010,
116
+ "left_ring_2_joint": 256,
117
+ "left_thumb_4_joint": 8,
118
+ "right_index_1_joint": 1018,
119
+ "right_little_1_joint": 243,
120
+ "right_little_2_joint": 12,
121
+ "right_middle_1_joint": 1411,
122
+ "right_middle_2_joint": 2,
123
+ "right_ring_1_joint": 440,
124
+ "right_ring_2_joint": 42,
125
+ "right_thumb_2_joint": 8239,
126
+ "right_thumb_4_joint": 21
127
+ },
128
+ "joint_limit_hard_counts": {
129
+ "left_index_1_joint": 74,
130
+ "left_index_2_joint": 92,
131
+ "left_little_1_joint": 61,
132
+ "left_little_2_joint": 163,
133
+ "left_middle_1_joint": 57,
134
+ "left_middle_2_joint": 190,
135
+ "left_ring_1_joint": 52,
136
+ "left_ring_2_joint": 209,
137
+ "left_thumb_4_joint": 1,
138
+ "right_index_1_joint": 56,
139
+ "right_index_2_joint": 2,
140
+ "right_little_1_joint": 21,
141
+ "right_little_2_joint": 1,
142
+ "right_middle_1_joint": 69,
143
+ "right_middle_2_joint": 7,
144
+ "right_ring_1_joint": 23,
145
+ "right_ring_2_joint": 3,
146
+ "right_thumb_2_joint": 105,
147
+ "right_thumb_4_joint": 5
148
+ },
149
+ "joint_limit_hard_episode_events": 233,
150
+ "finger_joint_limit_saturation_counts": {
151
+ "left_index_1_joint": 328,
152
+ "left_little_1_joint": 1101,
153
+ "left_middle_1_joint": 1048,
154
+ "left_ring_1_joint": 1139,
155
+ "left_ring_2_joint": 20,
156
+ "right_index_1_joint": 435,
157
+ "right_little_1_joint": 190,
158
+ "right_middle_1_joint": 571,
159
+ "right_ring_1_joint": 212,
160
+ "right_ring_2_joint": 8,
161
+ "right_thumb_2_joint": 4066,
162
+ "right_thumb_4_joint": 15
163
+ },
164
+ "joint_limit_max_excess_by_joint": {
165
+ "left_index_1_joint": 0.7408990859985352,
166
+ "left_index_2_joint": 0.05689479410648346,
167
+ "left_little_1_joint": 0.8472734093666077,
168
+ "left_little_2_joint": 0.11849485337734222,
169
+ "left_middle_1_joint": 0.8660928606987,
170
+ "left_middle_2_joint": 0.04905013367533684,
171
+ "left_ring_1_joint": 1.01592218875885,
172
+ "left_ring_2_joint": 0.5073496699333191,
173
+ "left_thumb_2_joint": 3.832954098470509e-05,
174
+ "left_thumb_3_joint": 2.089868939947337e-05,
175
+ "left_thumb_4_joint": 0.05195322632789612,
176
+ "right_index_1_joint": 1.4315546751022339,
177
+ "right_index_2_joint": 0.008559360168874264,
178
+ "right_little_1_joint": 0.9065614938735962,
179
+ "right_little_2_joint": 0.1250920295715332,
180
+ "right_middle_1_joint": 1.316969871520996,
181
+ "right_middle_2_joint": 0.015624959953129292,
182
+ "right_ring_1_joint": 0.8660089373588562,
183
+ "right_ring_2_joint": 0.237954244017601,
184
+ "right_thumb_1_joint": 2.7429845431470312e-05,
185
+ "right_thumb_2_joint": 0.6550672650337219,
186
+ "right_thumb_4_joint": 0.45664194226264954
187
+ },
188
+ "coverage": {
189
+ "hard_limit_source": {
190
+ "available": 50,
191
+ "unavailable": 0,
192
+ "unknown": 0
193
+ },
194
+ "joint_role_manifest": {
195
+ "available": 50,
196
+ "unavailable": 0,
197
+ "unknown": 0
198
+ }
199
+ },
200
+ "hard_limit_sources": {
201
+ "runtime.joint_pos_limits": 50
202
+ },
203
+ "soft_limit_sources": {
204
+ "runtime.soft_joint_pos_limits": 50
205
+ },
206
+ "unknown_runtime_joint_names": [],
207
+ "baseline_checks": {
208
+ "post_home": {
209
+ "checked": 50,
210
+ "hard_violation_episodes": 0
211
+ },
212
+ "post_reset": {
213
+ "checked": 50,
214
+ "hard_violation_episodes": 0
215
+ }
216
+ }
217
+ }
218
+ },
219
+ "robot_motion": {
220
+ "joint_acceleration_rms": 55.233971586929215,
221
+ "joint_effort_rms": 39.780327947681045,
222
+ "joint_jerk_rms": 5661.010807551933,
223
+ "joint_velocity_rms": 1.1129331363555133
224
+ },
225
+ "grasp": {
226
+ "kinematic_grasp_stability_index": 0.917715514401692,
227
+ "mean_kinematic_grasp_stability": 0.3542707828654008,
228
+ "per_object_best_kinematic_grasp_stability": {
229
+ "obj_011_banana_2": 0.7515396869379887,
230
+ "obj_013_apple_3": 0.7993548772220119,
231
+ "obj_013_apple_4": 0.25318013446872856
232
+ },
233
+ "grasp_gsi_diagnostics": {
234
+ "mean_anti_slip_score": 0.9224610118335081,
235
+ "mean_duration_score": 0.11937039371115706,
236
+ "mean_hold_score": 0.11227932592574326,
237
+ "mean_kinematic_stability_score": 0.07303490792355624,
238
+ "mean_legacy_gsi": 0.2810599944805178,
239
+ "mean_lift_score": 0.20077206432180641,
240
+ "mean_motion_stability_score": 0.10535359163582765
241
+ }
242
+ },
243
+ "tool": {
244
+ "tool_selection_accuracy": null,
245
+ "tool_switch_success_rate": null,
246
+ "tool_switch_total": 0,
247
+ "tool_switch_successful": 0
248
+ },
249
+ "runtime": {
250
+ "avg_steps": 1432.86,
251
+ "error_count": 0
252
+ }
253
+ }
254
+ }
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/Inv/summary.json ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100600000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 41,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1920,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.82,
50
+ "success_rate_ci": {
51
+ "mean": 0.82,
52
+ "ci_low": 0.7,
53
+ "ci_high": 0.92
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.975
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 22.533333333333335
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.72,
64
+ "hard_violation_rate": 0.26,
65
+ "drop_rate": 0.16,
66
+ "high_speed_violation_rate": 0.22,
67
+ "safety_violation_step_rate": 0.0003850808669820662
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.84,
74
+ "at_end_success_rate": 0.82,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.82
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.925,
81
+ "current_chain_depth_progress_score": 0.9533333333333334,
82
+ "latched_chain_depth_progress_score": 0.98,
83
+ "avg_current_chain_depth": 2.86,
84
+ "avg_latched_chain_depth": 2.94
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9479399032412894,
88
+ "expert_time_s": 21.333333333333332,
89
+ "expert_time_step": 1280,
90
+ "avg_steps_to_success": 1352,
91
+ "avg_policy_steps_to_success": 450.9512195121951,
92
+ "avg_policy_queries_to_success": 28.5609756097561
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0002888106502365497,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_013_apple_3": 5,
98
+ "drop_to_table_obj_013_apple_4": 5,
99
+ "high_speed_obj_013_apple_3": 8,
100
+ "high_speed_obj_013_apple_4": 3
101
+ },
102
+ "robot_constraint_violation_rate": 1.0,
103
+ "finger_joint_limit_saturation_rate": 0.98,
104
+ "joint_limit_active_step_rate": 0.1696968863461327,
105
+ "finger_joint_limit_saturation_step_rate": 0.08683573550445593,
106
+ "max_joint_limit_excess_rad": 1.3648929595947266,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "left_index_1_joint": 1407,
110
+ "left_index_2_joint": 50,
111
+ "left_little_1_joint": 1950,
112
+ "left_little_2_joint": 249,
113
+ "left_middle_1_joint": 1993,
114
+ "left_middle_2_joint": 201,
115
+ "left_ring_1_joint": 2155,
116
+ "left_ring_2_joint": 255,
117
+ "left_thumb_2_joint": 51,
118
+ "left_thumb_3_joint": 64,
119
+ "left_thumb_4_joint": 72,
120
+ "right_index_1_joint": 651,
121
+ "right_index_2_joint": 5,
122
+ "right_little_1_joint": 415,
123
+ "right_little_2_joint": 103,
124
+ "right_middle_1_joint": 997,
125
+ "right_middle_2_joint": 18,
126
+ "right_ring_1_joint": 486,
127
+ "right_ring_2_joint": 8,
128
+ "right_thumb_2_joint": 8711,
129
+ "right_thumb_4_joint": 96
130
+ },
131
+ "joint_limit_hard_counts": {
132
+ "left_index_1_joint": 84,
133
+ "left_index_2_joint": 102,
134
+ "left_little_1_joint": 58,
135
+ "left_little_2_joint": 185,
136
+ "left_middle_1_joint": 58,
137
+ "left_middle_2_joint": 181,
138
+ "left_ring_1_joint": 60,
139
+ "left_ring_2_joint": 203,
140
+ "left_thumb_2_joint": 3,
141
+ "left_thumb_3_joint": 3,
142
+ "left_thumb_4_joint": 8,
143
+ "right_index_1_joint": 35,
144
+ "right_index_2_joint": 5,
145
+ "right_little_1_joint": 21,
146
+ "right_little_2_joint": 7,
147
+ "right_middle_1_joint": 56,
148
+ "right_middle_2_joint": 20,
149
+ "right_ring_1_joint": 25,
150
+ "right_ring_2_joint": 3,
151
+ "right_thumb_2_joint": 101,
152
+ "right_thumb_4_joint": 11
153
+ },
154
+ "joint_limit_hard_episode_events": 244,
155
+ "finger_joint_limit_saturation_counts": {
156
+ "left_index_1_joint": 373,
157
+ "left_little_1_joint": 1201,
158
+ "left_middle_1_joint": 1098,
159
+ "left_ring_1_joint": 1307,
160
+ "left_thumb_3_joint": 40,
161
+ "left_thumb_4_joint": 8,
162
+ "right_index_1_joint": 240,
163
+ "right_little_1_joint": 335,
164
+ "right_little_2_joint": 46,
165
+ "right_middle_1_joint": 394,
166
+ "right_ring_1_joint": 311,
167
+ "right_thumb_2_joint": 4072,
168
+ "right_thumb_4_joint": 60
169
+ },
170
+ "joint_limit_max_excess_by_joint": {
171
+ "left_index_1_joint": 0.7143006920814514,
172
+ "left_index_2_joint": 0.039849646389484406,
173
+ "left_little_1_joint": 0.954632043838501,
174
+ "left_little_2_joint": 0.09940697252750397,
175
+ "left_middle_1_joint": 1.0076642036437988,
176
+ "left_middle_2_joint": 0.10911153256893158,
177
+ "left_ring_1_joint": 1.2215214967727661,
178
+ "left_ring_2_joint": 0.07283161580562592,
179
+ "left_thumb_2_joint": 0.143313467502594,
180
+ "left_thumb_3_joint": 0.6770806908607483,
181
+ "left_thumb_4_joint": 0.4676302969455719,
182
+ "right_index_1_joint": 0.4497883915901184,
183
+ "right_index_2_joint": 0.025190046057105064,
184
+ "right_little_1_joint": 1.3648929595947266,
185
+ "right_little_2_joint": 0.5094146728515625,
186
+ "right_middle_1_joint": 1.0405175685882568,
187
+ "right_middle_2_joint": 0.06751329451799393,
188
+ "right_ring_1_joint": 0.9371050596237183,
189
+ "right_ring_2_joint": 0.05560990422964096,
190
+ "right_thumb_1_joint": 0.0001693263475317508,
191
+ "right_thumb_2_joint": 0.5333144664764404,
192
+ "right_thumb_4_joint": 0.9461238384246826
193
+ },
194
+ "coverage": {
195
+ "hard_limit_source": {
196
+ "available": 50,
197
+ "unavailable": 0,
198
+ "unknown": 0
199
+ },
200
+ "joint_role_manifest": {
201
+ "available": 50,
202
+ "unavailable": 0,
203
+ "unknown": 0
204
+ }
205
+ },
206
+ "hard_limit_sources": {
207
+ "runtime.joint_pos_limits": 50
208
+ },
209
+ "soft_limit_sources": {
210
+ "runtime.soft_joint_pos_limits": 50
211
+ },
212
+ "unknown_runtime_joint_names": [],
213
+ "baseline_checks": {
214
+ "post_home": {
215
+ "checked": 50,
216
+ "hard_violation_episodes": 0
217
+ },
218
+ "post_reset": {
219
+ "checked": 50,
220
+ "hard_violation_episodes": 0
221
+ }
222
+ }
223
+ }
224
+ },
225
+ "robot_motion": {
226
+ "joint_acceleration_rms": 55.741338810073344,
227
+ "joint_effort_rms": 40.075257303755144,
228
+ "joint_jerk_rms": 5705.968277063157,
229
+ "joint_velocity_rms": 1.1200247985698504
230
+ },
231
+ "grasp": {
232
+ "kinematic_grasp_stability_index": 0.8599249776704512,
233
+ "mean_kinematic_grasp_stability": 0.3233615218836795,
234
+ "per_object_best_kinematic_grasp_stability": {
235
+ "obj_011_banana_2": 0.7154710423729606,
236
+ "obj_013_apple_3": 0.7588401840422256,
237
+ "obj_013_apple_4": 0.3075955913234023
238
+ },
239
+ "grasp_gsi_diagnostics": {
240
+ "mean_anti_slip_score": 0.8781346647463308,
241
+ "mean_duration_score": 0.14022377049931622,
242
+ "mean_hold_score": 0.12670388479021422,
243
+ "mean_kinematic_stability_score": 0.07284661048503344,
244
+ "mean_legacy_gsi": 0.2842983757487409,
245
+ "mean_lift_score": 0.2130525766656133,
246
+ "mean_motion_stability_score": 0.11866252469166144
247
+ }
248
+ },
249
+ "tool": {
250
+ "tool_selection_accuracy": null,
251
+ "tool_switch_success_rate": null,
252
+ "tool_switch_total": 0,
253
+ "tool_switch_successful": 0
254
+ },
255
+ "runtime": {
256
+ "avg_steps": 1454.24,
257
+ "error_count": 0
258
+ }
259
+ }
260
+ }
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/06_fruit_bowl_loading/None/summary.json ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100600000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 49,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1920,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.98,
50
+ "success_rate_ci": {
51
+ "mean": 0.98,
52
+ "ci_low": 0.94,
53
+ "ci_high": 1.0
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 1.0
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 22.944217687074808
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.68,
64
+ "hard_violation_rate": 0.32,
65
+ "drop_rate": 0.04,
66
+ "high_speed_violation_rate": 0.3,
67
+ "safety_violation_step_rate": 0.00037476937269372696
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 1.0,
74
+ "at_end_success_rate": 0.98,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.98
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.995,
81
+ "current_chain_depth_progress_score": 0.9933333333333333,
82
+ "latched_chain_depth_progress_score": 1.0,
83
+ "avg_current_chain_depth": 2.98,
84
+ "avg_latched_chain_depth": 3
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9317074673342799,
88
+ "expert_time_s": 21.333333333333332,
89
+ "expert_time_step": 1280,
90
+ "avg_steps_to_success": 1376.6530612244899,
91
+ "avg_policy_steps_to_success": 459.2244897959184,
92
+ "avg_policy_queries_to_success": 29.163265306122447
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.00030269833948339485,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_011_banana_2": 1,
98
+ "drop_to_table_obj_013_apple_4": 2,
99
+ "high_speed_obj_013_apple_3": 8,
100
+ "high_speed_obj_013_apple_4": 10
101
+ },
102
+ "robot_constraint_violation_rate": 1.0,
103
+ "finger_joint_limit_saturation_rate": 0.94,
104
+ "joint_limit_active_step_rate": 0.13954393450184502,
105
+ "finger_joint_limit_saturation_step_rate": 0.06332160977859778,
106
+ "max_joint_limit_excess_rad": 1.1884056329727173,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "left_index_1_joint": 1114,
110
+ "left_index_2_joint": 37,
111
+ "left_little_1_joint": 1456,
112
+ "left_little_2_joint": 259,
113
+ "left_middle_1_joint": 1556,
114
+ "left_middle_2_joint": 103,
115
+ "left_ring_1_joint": 1640,
116
+ "left_ring_2_joint": 253,
117
+ "left_thumb_2_joint": 76,
118
+ "left_thumb_3_joint": 36,
119
+ "left_thumb_4_joint": 17,
120
+ "right_index_1_joint": 150,
121
+ "right_little_1_joint": 58,
122
+ "right_middle_1_joint": 580,
123
+ "right_middle_2_joint": 9,
124
+ "right_ring_1_joint": 184,
125
+ "right_ring_2_joint": 15,
126
+ "right_thumb_2_joint": 7016,
127
+ "right_thumb_3_joint": 24,
128
+ "right_thumb_4_joint": 35
129
+ },
130
+ "joint_limit_hard_counts": {
131
+ "left_index_1_joint": 56,
132
+ "left_index_2_joint": 64,
133
+ "left_little_1_joint": 48,
134
+ "left_little_2_joint": 110,
135
+ "left_middle_1_joint": 52,
136
+ "left_middle_2_joint": 113,
137
+ "left_ring_1_joint": 50,
138
+ "left_ring_2_joint": 152,
139
+ "left_thumb_2_joint": 2,
140
+ "left_thumb_3_joint": 1,
141
+ "left_thumb_4_joint": 4,
142
+ "right_index_1_joint": 23,
143
+ "right_little_1_joint": 17,
144
+ "right_middle_1_joint": 37,
145
+ "right_middle_2_joint": 2,
146
+ "right_ring_1_joint": 22,
147
+ "right_ring_2_joint": 2,
148
+ "right_thumb_2_joint": 94,
149
+ "right_thumb_3_joint": 1,
150
+ "right_thumb_4_joint": 15
151
+ },
152
+ "joint_limit_hard_episode_events": 211,
153
+ "finger_joint_limit_saturation_counts": {
154
+ "left_index_1_joint": 411,
155
+ "left_little_1_joint": 925,
156
+ "left_little_2_joint": 15,
157
+ "left_middle_1_joint": 898,
158
+ "left_ring_1_joint": 955,
159
+ "left_ring_2_joint": 3,
160
+ "left_thumb_2_joint": 16,
161
+ "left_thumb_3_joint": 31,
162
+ "right_index_1_joint": 17,
163
+ "right_little_1_joint": 34,
164
+ "right_middle_1_joint": 73,
165
+ "right_ring_1_joint": 41,
166
+ "right_thumb_2_joint": 3019,
167
+ "right_thumb_3_joint": 19
168
+ },
169
+ "joint_limit_max_excess_by_joint": {
170
+ "left_index_1_joint": 0.9208499789237976,
171
+ "left_index_2_joint": 0.073265440762043,
172
+ "left_little_1_joint": 1.094996452331543,
173
+ "left_little_2_joint": 0.20538224279880524,
174
+ "left_middle_1_joint": 1.1240921020507812,
175
+ "left_middle_2_joint": 0.06215345114469528,
176
+ "left_ring_1_joint": 1.1884056329727173,
177
+ "left_ring_2_joint": 0.1506158858537674,
178
+ "left_thumb_2_joint": 0.2384660392999649,
179
+ "left_thumb_3_joint": 0.42807871103286743,
180
+ "left_thumb_4_joint": 0.09095409512519836,
181
+ "right_index_1_joint": 0.3336717188358307,
182
+ "right_index_2_joint": 2.430551376164658e-06,
183
+ "right_little_1_joint": 0.9055420756340027,
184
+ "right_little_2_joint": 1.7061773178284056e-05,
185
+ "right_middle_1_joint": 0.3945111036300659,
186
+ "right_middle_2_joint": 0.04110622778534889,
187
+ "right_ring_1_joint": 0.389538437128067,
188
+ "right_ring_2_joint": 0.04716918244957924,
189
+ "right_thumb_1_joint": 2.9913377147749998e-05,
190
+ "right_thumb_2_joint": 0.3231459856033325,
191
+ "right_thumb_3_joint": 0.6961715817451477,
192
+ "right_thumb_4_joint": 0.20877215266227722
193
+ },
194
+ "coverage": {
195
+ "hard_limit_source": {
196
+ "available": 50,
197
+ "unavailable": 0,
198
+ "unknown": 0
199
+ },
200
+ "joint_role_manifest": {
201
+ "available": 50,
202
+ "unavailable": 0,
203
+ "unknown": 0
204
+ }
205
+ },
206
+ "hard_limit_sources": {
207
+ "runtime.joint_pos_limits": 50
208
+ },
209
+ "soft_limit_sources": {
210
+ "runtime.soft_joint_pos_limits": 50
211
+ },
212
+ "unknown_runtime_joint_names": [],
213
+ "baseline_checks": {
214
+ "post_home": {
215
+ "checked": 50,
216
+ "hard_violation_episodes": 0
217
+ },
218
+ "post_reset": {
219
+ "checked": 50,
220
+ "hard_violation_episodes": 0
221
+ }
222
+ }
223
+ }
224
+ },
225
+ "robot_motion": {
226
+ "joint_acceleration_rms": 51.53312498291515,
227
+ "joint_effort_rms": 40.42516476339212,
228
+ "joint_jerk_rms": 5267.545594130939,
229
+ "joint_velocity_rms": 1.0944665228715633
230
+ },
231
+ "grasp": {
232
+ "kinematic_grasp_stability_index": 0.9408928541074613,
233
+ "mean_kinematic_grasp_stability": 0.35529683064606954,
234
+ "per_object_best_kinematic_grasp_stability": {
235
+ "obj_011_banana_2": 0.7895934169648127,
236
+ "obj_013_apple_3": 0.8468932790686428,
237
+ "obj_013_apple_4": 0.2380425143820733
238
+ },
239
+ "grasp_gsi_diagnostics": {
240
+ "mean_anti_slip_score": 0.9332318619446135,
241
+ "mean_duration_score": 0.11287887808723839,
242
+ "mean_hold_score": 0.10836426818629832,
243
+ "mean_kinematic_stability_score": 0.07053352596203905,
244
+ "mean_legacy_gsi": 0.27961324900181606,
245
+ "mean_lift_score": 0.19746681840132088,
246
+ "mean_motion_stability_score": 0.1014417135403216
247
+ }
248
+ },
249
+ "tool": {
250
+ "tool_selection_accuracy": null,
251
+ "tool_switch_success_rate": null,
252
+ "tool_switch_total": 0,
253
+ "tool_switch_successful": 0
254
+ },
255
+ "runtime": {
256
+ "avg_steps": 1387.52,
257
+ "error_count": 0
258
+ }
259
+ }
260
+ }
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Equi/summary.json ADDED
@@ -0,0 +1,277 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100700000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 46,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1806,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.92,
50
+ "success_rate_ci": {
51
+ "mean": 0.92,
52
+ "ci_low": 0.84,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.97
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 24.118840579710056
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.9,
64
+ "hard_violation_rate": 0.02,
65
+ "drop_rate": 0.0,
66
+ "high_speed_violation_rate": 0.02,
67
+ "safety_violation_step_rate": 1.355160450997398e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.96,
74
+ "at_end_success_rate": 0.92,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.92
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.96,
81
+ "current_chain_depth_progress_score": 0.98,
82
+ "latched_chain_depth_progress_score": 0.98,
83
+ "avg_current_chain_depth": 1.96,
84
+ "avg_latched_chain_depth": 1.96
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.833069786386294,
88
+ "expert_time_s": 20.05,
89
+ "expert_time_step": 1203,
90
+ "avg_steps_to_success": 1447.1304347826087,
91
+ "avg_policy_steps_to_success": 482.7173913043478,
92
+ "avg_policy_queries_to_success": 30.630434782608695
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 1.355160450997398e-05,
96
+ "violation_counts": {
97
+ "high_speed_obj_014_lemon_2": 1
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.96,
101
+ "joint_limit_active_step_rate": 0.09562012142237641,
102
+ "finger_joint_limit_saturation_step_rate": 0.024948503902862098,
103
+ "max_joint_limit_excess_rad": 1.0011656284332275,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "left_index_pitch_joint": 22,
107
+ "left_middle_dip_joint": 362,
108
+ "left_middle_pip_joint": 598,
109
+ "left_middle_pitch_joint": 27,
110
+ "left_pinky_mcp_joint": 1330,
111
+ "left_pinky_pip_joint": 21,
112
+ "left_ring_dip_joint": 692,
113
+ "left_ring_mcp_joint": 1,
114
+ "left_ring_pip_joint": 1109,
115
+ "left_thumb_dip_joint": 2,
116
+ "right_index_dip_joint": 47,
117
+ "right_index_pip_joint": 144,
118
+ "right_index_pitch_joint": 47,
119
+ "right_middle_dip_joint": 339,
120
+ "right_middle_pip_joint": 668,
121
+ "right_middle_pitch_joint": 103,
122
+ "right_pinky_dip_joint": 2,
123
+ "right_pinky_mcp_joint": 224,
124
+ "right_pinky_pip_joint": 869,
125
+ "right_ring_dip_joint": 726,
126
+ "right_ring_mcp_joint": 49,
127
+ "right_ring_pip_joint": 2390,
128
+ "right_thumb_dip_joint": 310,
129
+ "right_thumb_mcp_joint": 2320,
130
+ "right_thumb_pip_joint": 17
131
+ },
132
+ "joint_limit_hard_counts": {
133
+ "left_index_dip_joint": 2,
134
+ "left_index_pip_joint": 1,
135
+ "left_index_pitch_joint": 1,
136
+ "left_middle_dip_joint": 155,
137
+ "left_middle_pip_joint": 118,
138
+ "left_middle_pitch_joint": 1,
139
+ "left_pinky_mcp_joint": 130,
140
+ "left_pinky_pip_joint": 2,
141
+ "left_ring_dip_joint": 219,
142
+ "left_ring_mcp_joint": 1,
143
+ "left_ring_pip_joint": 151,
144
+ "left_thumb_dip_joint": 5,
145
+ "right_index_dip_joint": 24,
146
+ "right_index_pip_joint": 16,
147
+ "right_index_pitch_joint": 6,
148
+ "right_middle_dip_joint": 142,
149
+ "right_middle_pip_joint": 81,
150
+ "right_middle_pitch_joint": 34,
151
+ "right_pinky_dip_joint": 1,
152
+ "right_pinky_mcp_joint": 23,
153
+ "right_pinky_pip_joint": 32,
154
+ "right_ring_dip_joint": 93,
155
+ "right_ring_mcp_joint": 9,
156
+ "right_ring_pip_joint": 100,
157
+ "right_thumb_dip_joint": 11,
158
+ "right_thumb_mcp_joint": 67,
159
+ "right_thumb_pip_joint": 1
160
+ },
161
+ "joint_limit_hard_episode_events": 393,
162
+ "finger_joint_limit_saturation_counts": {
163
+ "left_middle_pip_joint": 37,
164
+ "left_pinky_mcp_joint": 303,
165
+ "left_ring_dip_joint": 8,
166
+ "left_ring_pip_joint": 174,
167
+ "right_index_pip_joint": 74,
168
+ "right_index_pitch_joint": 16,
169
+ "right_middle_pip_joint": 186,
170
+ "right_middle_pitch_joint": 6,
171
+ "right_pinky_mcp_joint": 43,
172
+ "right_pinky_pip_joint": 527,
173
+ "right_ring_dip_joint": 69,
174
+ "right_ring_mcp_joint": 24,
175
+ "right_ring_pip_joint": 1221,
176
+ "right_thumb_dip_joint": 261,
177
+ "right_thumb_mcp_joint": 13
178
+ },
179
+ "joint_limit_max_excess_by_joint": {
180
+ "left_index_dip_joint": 0.006867041811347008,
181
+ "left_index_pip_joint": 0.005910896696150303,
182
+ "left_index_pitch_joint": 0.1051860898733139,
183
+ "left_middle_dip_joint": 0.10611969232559204,
184
+ "left_middle_pip_joint": 0.27826938033103943,
185
+ "left_middle_pitch_joint": 0.06385832279920578,
186
+ "left_pinky_mcp_joint": 0.5886069536209106,
187
+ "left_pinky_pip_joint": 0.11755232512950897,
188
+ "left_ring_dip_joint": 0.18967007100582123,
189
+ "left_ring_mcp_joint": 0.011387924663722515,
190
+ "left_ring_pip_joint": 0.4397903382778168,
191
+ "left_thumb_dip_joint": 0.017623206600546837,
192
+ "left_thumb_yaw_joint": 3.753419514396228e-05,
193
+ "right_index_dip_joint": 0.09440450370311737,
194
+ "right_index_pip_joint": 0.3292716145515442,
195
+ "right_index_pitch_joint": 0.27372050285339355,
196
+ "right_middle_dip_joint": 0.18065284192562103,
197
+ "right_middle_pip_joint": 0.5433264374732971,
198
+ "right_middle_pitch_joint": 0.20356141030788422,
199
+ "right_pinky_dip_joint": 0.013808801770210266,
200
+ "right_pinky_mcp_joint": 0.4709426462650299,
201
+ "right_pinky_pip_joint": 0.821548581123352,
202
+ "right_ring_dip_joint": 0.3088793456554413,
203
+ "right_ring_mcp_joint": 0.3128192722797394,
204
+ "right_ring_pip_joint": 0.7752983570098877,
205
+ "right_thumb_dip_joint": 1.0011656284332275,
206
+ "right_thumb_mcp_joint": 0.26057249307632446,
207
+ "right_thumb_pip_joint": 0.1779090166091919,
208
+ "right_thumb_yaw_joint": 4.204039214528166e-05
209
+ },
210
+ "coverage": {
211
+ "hard_limit_source": {
212
+ "available": 50,
213
+ "unavailable": 0,
214
+ "unknown": 0
215
+ },
216
+ "joint_role_manifest": {
217
+ "available": 50,
218
+ "unavailable": 0,
219
+ "unknown": 0
220
+ }
221
+ },
222
+ "hard_limit_sources": {
223
+ "runtime.joint_pos_limits": 50
224
+ },
225
+ "soft_limit_sources": {
226
+ "runtime.soft_joint_pos_limits": 50
227
+ },
228
+ "unknown_runtime_joint_names": [],
229
+ "baseline_checks": {
230
+ "post_home": {
231
+ "checked": 50,
232
+ "hard_violation_episodes": 0
233
+ },
234
+ "post_reset": {
235
+ "checked": 50,
236
+ "hard_violation_episodes": 0
237
+ }
238
+ }
239
+ }
240
+ },
241
+ "robot_motion": {
242
+ "joint_acceleration_rms": 85.84063247352294,
243
+ "joint_effort_rms": 22.300521247227785,
244
+ "joint_jerk_rms": 8819.261320647254,
245
+ "joint_velocity_rms": 1.1406098236639461
246
+ },
247
+ "grasp": {
248
+ "kinematic_grasp_stability_index": 0.8874566383701085,
249
+ "mean_kinematic_grasp_stability": 0.5441457554890654,
250
+ "per_object_best_kinematic_grasp_stability": {
251
+ "obj_014_lemon_2": 0.6418838579922247,
252
+ "obj_014_lemon_3": 0.6336913949197229,
253
+ "obj_017_orange_4": 0.44473737235688254,
254
+ "obj_017_orange_5": 0.3985262463441392
255
+ },
256
+ "grasp_gsi_diagnostics": {
257
+ "mean_anti_slip_score": 0.7049488200176083,
258
+ "mean_duration_score": 0.41473966572592574,
259
+ "mean_hold_score": 0.363759583856167,
260
+ "mean_kinematic_stability_score": 0.13421336944986265,
261
+ "mean_legacy_gsi": 0.44210523039221,
262
+ "mean_lift_score": 0.4348382953736203,
263
+ "mean_motion_stability_score": 0.3539467577839904
264
+ }
265
+ },
266
+ "tool": {
267
+ "tool_selection_accuracy": null,
268
+ "tool_switch_success_rate": null,
269
+ "tool_switch_total": 0,
270
+ "tool_switch_successful": 0
271
+ },
272
+ "runtime": {
273
+ "avg_steps": 1475.84,
274
+ "error_count": 0
275
+ }
276
+ }
277
+ }
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Full/summary.json ADDED
@@ -0,0 +1,261 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 200700000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 47,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1806,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.94,
50
+ "success_rate_ci": {
51
+ "mean": 0.94,
52
+ "ci_low": 0.86,
53
+ "ci_high": 1.0
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.985
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.591843971631146
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.88,
64
+ "hard_violation_rate": 0.1,
65
+ "drop_rate": 0.04,
66
+ "high_speed_violation_rate": 0.06,
67
+ "safety_violation_step_rate": 6.949560092846123e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.98,
74
+ "at_end_success_rate": 0.94,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.94
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.975,
81
+ "current_chain_depth_progress_score": 0.99,
82
+ "latched_chain_depth_progress_score": 0.99,
83
+ "avg_current_chain_depth": 1.98,
84
+ "avg_latched_chain_depth": 1.98
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.8508314326329129,
88
+ "expert_time_s": 20.05,
89
+ "expert_time_step": 1203,
90
+ "avg_steps_to_success": 1415.5106382978724,
91
+ "avg_policy_steps_to_success": 472.1063829787234,
92
+ "avg_policy_queries_to_success": 29.95744680851064
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 6.949560092846123e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_014_lemon_2": 1,
98
+ "drop_to_table_obj_017_orange_5": 1,
99
+ "high_speed_obj_014_lemon_2": 3
100
+ },
101
+ "robot_constraint_violation_rate": 1.0,
102
+ "finger_joint_limit_saturation_rate": 0.96,
103
+ "joint_limit_active_step_rate": 0.10479936620011954,
104
+ "finger_joint_limit_saturation_step_rate": 0.020654092595938676,
105
+ "max_joint_limit_excess_rad": 0.7954468131065369,
106
+ "robot_constraint_diagnostics": {
107
+ "joint_limit_soft_counts": {
108
+ "left_middle_dip_joint": 321,
109
+ "left_middle_pip_joint": 425,
110
+ "left_pinky_dip_joint": 13,
111
+ "left_pinky_mcp_joint": 1551,
112
+ "left_pinky_pip_joint": 19,
113
+ "left_ring_dip_joint": 792,
114
+ "left_ring_pip_joint": 1142,
115
+ "right_index_dip_joint": 13,
116
+ "right_index_pip_joint": 90,
117
+ "right_index_pitch_joint": 18,
118
+ "right_middle_dip_joint": 190,
119
+ "right_middle_pip_joint": 405,
120
+ "right_middle_pitch_joint": 132,
121
+ "right_pinky_mcp_joint": 106,
122
+ "right_pinky_pip_joint": 626,
123
+ "right_ring_dip_joint": 504,
124
+ "right_ring_mcp_joint": 39,
125
+ "right_ring_pip_joint": 2144,
126
+ "right_thumb_dip_joint": 213,
127
+ "right_thumb_mcp_joint": 2819
128
+ },
129
+ "joint_limit_hard_counts": {
130
+ "left_middle_dip_joint": 178,
131
+ "left_middle_pip_joint": 155,
132
+ "left_pinky_dip_joint": 1,
133
+ "left_pinky_mcp_joint": 132,
134
+ "left_pinky_pip_joint": 1,
135
+ "left_ring_dip_joint": 231,
136
+ "left_ring_pip_joint": 168,
137
+ "right_index_dip_joint": 11,
138
+ "right_index_pip_joint": 9,
139
+ "right_index_pitch_joint": 6,
140
+ "right_middle_dip_joint": 102,
141
+ "right_middle_pip_joint": 64,
142
+ "right_middle_pitch_joint": 41,
143
+ "right_pinky_mcp_joint": 22,
144
+ "right_pinky_pip_joint": 26,
145
+ "right_ring_dip_joint": 55,
146
+ "right_ring_mcp_joint": 17,
147
+ "right_ring_pip_joint": 95,
148
+ "right_thumb_dip_joint": 6,
149
+ "right_thumb_mcp_joint": 89
150
+ },
151
+ "joint_limit_hard_episode_events": 434,
152
+ "finger_joint_limit_saturation_counts": {
153
+ "left_middle_pip_joint": 20,
154
+ "left_pinky_mcp_joint": 271,
155
+ "left_pinky_pip_joint": 3,
156
+ "left_ring_dip_joint": 16,
157
+ "left_ring_pip_joint": 92,
158
+ "right_middle_pip_joint": 78,
159
+ "right_middle_pitch_joint": 6,
160
+ "right_pinky_mcp_joint": 2,
161
+ "right_pinky_pip_joint": 388,
162
+ "right_ring_dip_joint": 59,
163
+ "right_ring_pip_joint": 1012,
164
+ "right_thumb_dip_joint": 188,
165
+ "right_thumb_mcp_joint": 37
166
+ },
167
+ "joint_limit_max_excess_by_joint": {
168
+ "left_index_dip_joint": 6.348179886117578e-05,
169
+ "left_index_pip_joint": 0.0020269800443202257,
170
+ "left_middle_dip_joint": 0.10114391893148422,
171
+ "left_middle_pip_joint": 0.3214621841907501,
172
+ "left_pinky_dip_joint": 0.15028798580169678,
173
+ "left_pinky_mcp_joint": 0.34777072072029114,
174
+ "left_pinky_pip_joint": 0.2511374354362488,
175
+ "left_ring_dip_joint": 0.2365267127752304,
176
+ "left_ring_pip_joint": 0.3900928199291229,
177
+ "left_thumb_dip_joint": 0.0043436684645712376,
178
+ "left_thumb_yaw_joint": 2.941852835647296e-05,
179
+ "right_index_dip_joint": 0.04829217493534088,
180
+ "right_index_pip_joint": 0.11951853334903717,
181
+ "right_index_pitch_joint": 0.10652457177639008,
182
+ "right_middle_dip_joint": 0.11580635607242584,
183
+ "right_middle_pip_joint": 0.45964398980140686,
184
+ "right_middle_pitch_joint": 0.19317692518234253,
185
+ "right_pinky_mcp_joint": 0.18345560133457184,
186
+ "right_pinky_pip_joint": 0.6609310507774353,
187
+ "right_ring_dip_joint": 0.2952408492565155,
188
+ "right_ring_mcp_joint": 0.09753109514713287,
189
+ "right_ring_pip_joint": 0.5822908878326416,
190
+ "right_thumb_dip_joint": 0.7954468131065369,
191
+ "right_thumb_mcp_joint": 0.15592491626739502,
192
+ "right_thumb_yaw_joint": 3.000023207277991e-05
193
+ },
194
+ "coverage": {
195
+ "hard_limit_source": {
196
+ "available": 50,
197
+ "unavailable": 0,
198
+ "unknown": 0
199
+ },
200
+ "joint_role_manifest": {
201
+ "available": 50,
202
+ "unavailable": 0,
203
+ "unknown": 0
204
+ }
205
+ },
206
+ "hard_limit_sources": {
207
+ "runtime.joint_pos_limits": 50
208
+ },
209
+ "soft_limit_sources": {
210
+ "runtime.soft_joint_pos_limits": 50
211
+ },
212
+ "unknown_runtime_joint_names": [],
213
+ "baseline_checks": {
214
+ "post_home": {
215
+ "checked": 50,
216
+ "hard_violation_episodes": 0
217
+ },
218
+ "post_reset": {
219
+ "checked": 50,
220
+ "hard_violation_episodes": 0
221
+ }
222
+ }
223
+ }
224
+ },
225
+ "robot_motion": {
226
+ "joint_acceleration_rms": 87.20556197498117,
227
+ "joint_effort_rms": 22.497489873570434,
228
+ "joint_jerk_rms": 8923.907545742004,
229
+ "joint_velocity_rms": 1.1410834881321135
230
+ },
231
+ "grasp": {
232
+ "kinematic_grasp_stability_index": 0.829160622727524,
233
+ "mean_kinematic_grasp_stability": 0.455741119366631,
234
+ "per_object_best_kinematic_grasp_stability": {
235
+ "obj_014_lemon_2": 0.5899051901984458,
236
+ "obj_014_lemon_3": 0.556119748121484,
237
+ "obj_017_orange_4": 0.4506548068166049,
238
+ "obj_017_orange_5": 0.44217670134222126
239
+ },
240
+ "grasp_gsi_diagnostics": {
241
+ "mean_anti_slip_score": 0.6785548370288587,
242
+ "mean_duration_score": 0.41230198498144843,
243
+ "mean_hold_score": 0.37448700918809347,
244
+ "mean_kinematic_stability_score": 0.11129744465709987,
245
+ "mean_legacy_gsi": 0.44189050570784744,
246
+ "mean_lift_score": 0.4467868575320363,
247
+ "mean_motion_stability_score": 0.36269799729811947
248
+ }
249
+ },
250
+ "tool": {
251
+ "tool_selection_accuracy": null,
252
+ "tool_switch_success_rate": null,
253
+ "tool_switch_total": 0,
254
+ "tool_switch_successful": 0
255
+ },
256
+ "runtime": {
257
+ "avg_steps": 1438.94,
258
+ "error_count": 0
259
+ }
260
+ }
261
+ }
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/07_citrus_plate_loading/Inv/summary.json ADDED
@@ -0,0 +1,273 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100700000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 44,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1806,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.88,
50
+ "success_rate_ci": {
51
+ "mean": 0.88,
52
+ "ci_low": 0.78,
53
+ "ci_high": 0.96
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.96
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.873484848484775
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.88,
64
+ "hard_violation_rate": 0.02,
65
+ "drop_rate": 0.02,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 1.353876147410035e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.94,
74
+ "at_end_success_rate": 0.88,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.88
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.945,
81
+ "current_chain_depth_progress_score": 0.97,
82
+ "latched_chain_depth_progress_score": 0.97,
83
+ "avg_current_chain_depth": 1.94,
84
+ "avg_latched_chain_depth": 1.94
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.84131999681249,
88
+ "expert_time_s": 20.05,
89
+ "expert_time_step": 1203,
90
+ "avg_steps_to_success": 1432.409090909091,
91
+ "avg_policy_steps_to_success": 477.72727272727275,
92
+ "avg_policy_queries_to_success": 30.386363636363637
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 1.353876147410035e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_017_orange_5": 1
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 1.0,
101
+ "joint_limit_active_step_rate": 0.11732690693455362,
102
+ "finger_joint_limit_saturation_step_rate": 0.029324957352901356,
103
+ "max_joint_limit_excess_rad": 1.394063115119934,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "left_index_dip_joint": 6,
107
+ "left_index_pip_joint": 6,
108
+ "left_middle_dip_joint": 305,
109
+ "left_middle_pip_joint": 475,
110
+ "left_pinky_mcp_joint": 1447,
111
+ "left_pinky_pip_joint": 31,
112
+ "left_ring_dip_joint": 684,
113
+ "left_ring_pip_joint": 1034,
114
+ "right_index_dip_joint": 73,
115
+ "right_index_pip_joint": 321,
116
+ "right_index_pitch_joint": 194,
117
+ "right_middle_dip_joint": 228,
118
+ "right_middle_pip_joint": 720,
119
+ "right_middle_pitch_joint": 333,
120
+ "right_pinky_mcp_joint": 455,
121
+ "right_pinky_pip_joint": 960,
122
+ "right_ring_dip_joint": 697,
123
+ "right_ring_mcp_joint": 315,
124
+ "right_ring_pip_joint": 2653,
125
+ "right_thumb_dip_joint": 441,
126
+ "right_thumb_mcp_joint": 3491,
127
+ "right_thumb_pip_joint": 20,
128
+ "right_thumb_yaw_joint": 71
129
+ },
130
+ "joint_limit_hard_counts": {
131
+ "left_index_dip_joint": 5,
132
+ "left_index_pip_joint": 5,
133
+ "left_middle_dip_joint": 143,
134
+ "left_middle_pip_joint": 111,
135
+ "left_pinky_mcp_joint": 101,
136
+ "left_pinky_pip_joint": 2,
137
+ "left_ring_dip_joint": 233,
138
+ "left_ring_pip_joint": 177,
139
+ "right_index_dip_joint": 23,
140
+ "right_index_pip_joint": 25,
141
+ "right_index_pitch_joint": 19,
142
+ "right_middle_dip_joint": 134,
143
+ "right_middle_pip_joint": 100,
144
+ "right_middle_pitch_joint": 62,
145
+ "right_pinky_dip_joint": 2,
146
+ "right_pinky_mcp_joint": 31,
147
+ "right_pinky_pip_joint": 39,
148
+ "right_ring_dip_joint": 87,
149
+ "right_ring_mcp_joint": 15,
150
+ "right_ring_pip_joint": 119,
151
+ "right_thumb_dip_joint": 13,
152
+ "right_thumb_mcp_joint": 93,
153
+ "right_thumb_pip_joint": 1,
154
+ "right_thumb_yaw_joint": 3
155
+ },
156
+ "joint_limit_hard_episode_events": 446,
157
+ "finger_joint_limit_saturation_counts": {
158
+ "left_middle_pip_joint": 86,
159
+ "left_pinky_mcp_joint": 216,
160
+ "left_pinky_pip_joint": 14,
161
+ "left_ring_dip_joint": 3,
162
+ "left_ring_pip_joint": 173,
163
+ "right_index_pip_joint": 156,
164
+ "right_index_pitch_joint": 147,
165
+ "right_middle_pip_joint": 236,
166
+ "right_middle_pitch_joint": 165,
167
+ "right_pinky_mcp_joint": 272,
168
+ "right_pinky_pip_joint": 595,
169
+ "right_ring_dip_joint": 66,
170
+ "right_ring_mcp_joint": 248,
171
+ "right_ring_pip_joint": 1417,
172
+ "right_thumb_dip_joint": 384,
173
+ "right_thumb_mcp_joint": 20,
174
+ "right_thumb_pip_joint": 3,
175
+ "right_thumb_yaw_joint": 35
176
+ },
177
+ "joint_limit_max_excess_by_joint": {
178
+ "left_index_dip_joint": 0.03131571412086487,
179
+ "left_index_pip_joint": 0.029112495481967926,
180
+ "left_middle_dip_joint": 0.11480570584535599,
181
+ "left_middle_pip_joint": 0.593149721622467,
182
+ "left_middle_pitch_joint": 0.00048358284402638674,
183
+ "left_pinky_mcp_joint": 0.5875157713890076,
184
+ "left_pinky_pip_joint": 0.31576746702194214,
185
+ "left_ring_dip_joint": 0.1824529618024826,
186
+ "left_ring_pip_joint": 0.5797896981239319,
187
+ "left_thumb_dip_joint": 2.431639359201654e-06,
188
+ "left_thumb_yaw_joint": 3.094927888014354e-05,
189
+ "right_index_dip_joint": 0.08236277103424072,
190
+ "right_index_pip_joint": 0.37338656187057495,
191
+ "right_index_pitch_joint": 0.6654715538024902,
192
+ "right_middle_dip_joint": 0.09381233900785446,
193
+ "right_middle_pip_joint": 0.5180755257606506,
194
+ "right_middle_pitch_joint": 0.8524304032325745,
195
+ "right_pinky_dip_joint": 0.005832630209624767,
196
+ "right_pinky_mcp_joint": 1.139074444770813,
197
+ "right_pinky_pip_joint": 0.8547611236572266,
198
+ "right_ring_dip_joint": 0.3315008282661438,
199
+ "right_ring_mcp_joint": 0.9743439555168152,
200
+ "right_ring_pip_joint": 1.394063115119934,
201
+ "right_thumb_dip_joint": 1.0448678731918335,
202
+ "right_thumb_mcp_joint": 0.1458820253610611,
203
+ "right_thumb_pip_joint": 0.16270720958709717,
204
+ "right_thumb_yaw_joint": 0.47444045543670654
205
+ },
206
+ "coverage": {
207
+ "hard_limit_source": {
208
+ "available": 50,
209
+ "unavailable": 0,
210
+ "unknown": 0
211
+ },
212
+ "joint_role_manifest": {
213
+ "available": 50,
214
+ "unavailable": 0,
215
+ "unknown": 0
216
+ }
217
+ },
218
+ "hard_limit_sources": {
219
+ "runtime.joint_pos_limits": 50
220
+ },
221
+ "soft_limit_sources": {
222
+ "runtime.soft_joint_pos_limits": 50
223
+ },
224
+ "unknown_runtime_joint_names": [],
225
+ "baseline_checks": {
226
+ "post_home": {
227
+ "checked": 50,
228
+ "hard_violation_episodes": 0
229
+ },
230
+ "post_reset": {
231
+ "checked": 50,
232
+ "hard_violation_episodes": 0
233
+ }
234
+ }
235
+ }
236
+ },
237
+ "robot_motion": {
238
+ "joint_acceleration_rms": 83.49954972466871,
239
+ "joint_effort_rms": 23.026458333197542,
240
+ "joint_jerk_rms": 8585.782494045687,
241
+ "joint_velocity_rms": 1.1081305227496099
242
+ },
243
+ "grasp": {
244
+ "kinematic_grasp_stability_index": 0.8350687804084914,
245
+ "mean_kinematic_grasp_stability": 0.4673601905015906,
246
+ "per_object_best_kinematic_grasp_stability": {
247
+ "obj_014_lemon_2": 0.6352411315412713,
248
+ "obj_014_lemon_3": 0.5465044762197823,
249
+ "obj_017_orange_4": 0.4650784524448455,
250
+ "obj_017_orange_5": 0.40891593517068325
251
+ },
252
+ "grasp_gsi_diagnostics": {
253
+ "mean_anti_slip_score": 0.6819452497307943,
254
+ "mean_duration_score": 0.4224431980736668,
255
+ "mean_hold_score": 0.3761536656821266,
256
+ "mean_kinematic_stability_score": 0.11892910578496874,
257
+ "mean_legacy_gsi": 0.44494982180474807,
258
+ "mean_lift_score": 0.44759831574456665,
259
+ "mean_motion_stability_score": 0.3633369977292153
260
+ }
261
+ },
262
+ "tool": {
263
+ "tool_selection_accuracy": null,
264
+ "tool_switch_success_rate": null,
265
+ "tool_switch_total": 0,
266
+ "tool_switch_successful": 0
267
+ },
268
+ "runtime": {
269
+ "avg_steps": 1477.24,
270
+ "error_count": 0
271
+ }
272
+ }
273
+ }
me-dex-1-5/per_task_eval/07_citrus_plate_loading/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/07_citrus_plate_loading/None/summary.json ADDED
@@ -0,0 +1,275 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100700000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 46,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1806,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.92,
50
+ "success_rate_ci": {
51
+ "mean": 0.92,
52
+ "ci_low": 0.84,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.985
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 24.22971014492744
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.9,
64
+ "hard_violation_rate": 0.02,
65
+ "drop_rate": 0.0,
66
+ "high_speed_violation_rate": 0.02,
67
+ "safety_violation_step_rate": 1.349564090798672e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.98,
74
+ "at_end_success_rate": 0.92,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.92
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.97,
81
+ "current_chain_depth_progress_score": 0.99,
82
+ "latched_chain_depth_progress_score": 0.99,
83
+ "avg_current_chain_depth": 1.98,
84
+ "avg_latched_chain_depth": 1.98
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.830376225951938,
88
+ "expert_time_s": 20.05,
89
+ "expert_time_step": 1203,
90
+ "avg_steps_to_success": 1453.7826086956522,
91
+ "avg_policy_steps_to_success": 485.0217391304348,
92
+ "avg_policy_queries_to_success": 30.695652173913043
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 1.349564090798672e-05,
96
+ "violation_counts": {
97
+ "high_speed_obj_014_lemon_3": 1
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.94,
101
+ "joint_limit_active_step_rate": 0.09514426840130638,
102
+ "finger_joint_limit_saturation_step_rate": 0.02547977003427893,
103
+ "max_joint_limit_excess_rad": 0.9545888900756836,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "left_index_dip_joint": 12,
107
+ "left_index_pip_joint": 26,
108
+ "left_index_pitch_joint": 55,
109
+ "left_middle_dip_joint": 290,
110
+ "left_middle_pip_joint": 598,
111
+ "left_middle_pitch_joint": 68,
112
+ "left_pinky_dip_joint": 11,
113
+ "left_pinky_mcp_joint": 1253,
114
+ "left_pinky_pip_joint": 27,
115
+ "left_ring_dip_joint": 620,
116
+ "left_ring_mcp_joint": 4,
117
+ "left_ring_pip_joint": 1259,
118
+ "left_thumb_dip_joint": 27,
119
+ "right_index_dip_joint": 18,
120
+ "right_index_pip_joint": 26,
121
+ "right_index_pitch_joint": 11,
122
+ "right_middle_dip_joint": 316,
123
+ "right_middle_pip_joint": 513,
124
+ "right_middle_pitch_joint": 104,
125
+ "right_pinky_mcp_joint": 263,
126
+ "right_pinky_pip_joint": 547,
127
+ "right_ring_dip_joint": 548,
128
+ "right_ring_pip_joint": 2229,
129
+ "right_thumb_dip_joint": 186,
130
+ "right_thumb_mcp_joint": 2393
131
+ },
132
+ "joint_limit_hard_counts": {
133
+ "left_index_dip_joint": 11,
134
+ "left_index_pip_joint": 7,
135
+ "left_index_pitch_joint": 1,
136
+ "left_middle_dip_joint": 127,
137
+ "left_middle_pip_joint": 102,
138
+ "left_middle_pitch_joint": 1,
139
+ "left_pinky_dip_joint": 1,
140
+ "left_pinky_mcp_joint": 129,
141
+ "left_pinky_pip_joint": 3,
142
+ "left_ring_dip_joint": 182,
143
+ "left_ring_mcp_joint": 3,
144
+ "left_ring_pip_joint": 128,
145
+ "left_thumb_dip_joint": 3,
146
+ "left_thumb_mcp_joint": 1,
147
+ "right_index_dip_joint": 14,
148
+ "right_index_pip_joint": 19,
149
+ "right_index_pitch_joint": 9,
150
+ "right_middle_dip_joint": 108,
151
+ "right_middle_pip_joint": 71,
152
+ "right_middle_pitch_joint": 50,
153
+ "right_pinky_mcp_joint": 30,
154
+ "right_pinky_pip_joint": 26,
155
+ "right_ring_dip_joint": 72,
156
+ "right_ring_mcp_joint": 6,
157
+ "right_ring_pip_joint": 92,
158
+ "right_thumb_dip_joint": 12,
159
+ "right_thumb_mcp_joint": 73
160
+ },
161
+ "joint_limit_hard_episode_events": 394,
162
+ "finger_joint_limit_saturation_counts": {
163
+ "left_index_pitch_joint": 18,
164
+ "left_middle_pip_joint": 115,
165
+ "left_middle_pitch_joint": 40,
166
+ "left_pinky_mcp_joint": 215,
167
+ "left_ring_dip_joint": 16,
168
+ "left_ring_pip_joint": 264,
169
+ "right_middle_pip_joint": 63,
170
+ "right_pinky_mcp_joint": 21,
171
+ "right_pinky_pip_joint": 310,
172
+ "right_ring_dip_joint": 38,
173
+ "right_ring_pip_joint": 1248,
174
+ "right_thumb_dip_joint": 134,
175
+ "right_thumb_mcp_joint": 26
176
+ },
177
+ "joint_limit_max_excess_by_joint": {
178
+ "left_index_dip_joint": 0.031983040273189545,
179
+ "left_index_pip_joint": 0.196244478225708,
180
+ "left_index_pitch_joint": 0.24060268700122833,
181
+ "left_middle_dip_joint": 0.10716956108808517,
182
+ "left_middle_pip_joint": 0.4393370747566223,
183
+ "left_middle_pitch_joint": 0.2218441516160965,
184
+ "left_pinky_dip_joint": 0.05261504650115967,
185
+ "left_pinky_mcp_joint": 0.4358605742454529,
186
+ "left_pinky_pip_joint": 0.1435207575559616,
187
+ "left_ring_dip_joint": 0.19275173544883728,
188
+ "left_ring_mcp_joint": 0.021673966199159622,
189
+ "left_ring_pip_joint": 0.6579980254173279,
190
+ "left_thumb_dip_joint": 0.08365371078252792,
191
+ "left_thumb_mcp_joint": 0.007243730127811432,
192
+ "left_thumb_yaw_joint": 2.044576103799045e-05,
193
+ "right_index_dip_joint": 0.025403402745723724,
194
+ "right_index_pip_joint": 0.04487897828221321,
195
+ "right_index_pitch_joint": 0.044391170144081116,
196
+ "right_middle_dip_joint": 0.12465567886829376,
197
+ "right_middle_pip_joint": 0.3078276813030243,
198
+ "right_middle_pitch_joint": 0.08199431747198105,
199
+ "right_pinky_mcp_joint": 0.3023759722709656,
200
+ "right_pinky_pip_joint": 0.9545888900756836,
201
+ "right_ring_dip_joint": 0.2673702836036682,
202
+ "right_ring_mcp_joint": 0.008058561012148857,
203
+ "right_ring_pip_joint": 0.7957926988601685,
204
+ "right_thumb_dip_joint": 0.799296498298645,
205
+ "right_thumb_mcp_joint": 0.36689549684524536,
206
+ "right_thumb_yaw_joint": 3.722383553395048e-05
207
+ },
208
+ "coverage": {
209
+ "hard_limit_source": {
210
+ "available": 50,
211
+ "unavailable": 0,
212
+ "unknown": 0
213
+ },
214
+ "joint_role_manifest": {
215
+ "available": 50,
216
+ "unavailable": 0,
217
+ "unknown": 0
218
+ }
219
+ },
220
+ "hard_limit_sources": {
221
+ "runtime.joint_pos_limits": 50
222
+ },
223
+ "soft_limit_sources": {
224
+ "runtime.soft_joint_pos_limits": 50
225
+ },
226
+ "unknown_runtime_joint_names": [],
227
+ "baseline_checks": {
228
+ "post_home": {
229
+ "checked": 50,
230
+ "hard_violation_episodes": 0
231
+ },
232
+ "post_reset": {
233
+ "checked": 50,
234
+ "hard_violation_episodes": 0
235
+ }
236
+ }
237
+ }
238
+ },
239
+ "robot_motion": {
240
+ "joint_acceleration_rms": 80.89633958251363,
241
+ "joint_effort_rms": 22.337670394389107,
242
+ "joint_jerk_rms": 8257.898614465052,
243
+ "joint_velocity_rms": 1.0830466332399158
244
+ },
245
+ "grasp": {
246
+ "kinematic_grasp_stability_index": 0.857198721323545,
247
+ "mean_kinematic_grasp_stability": 0.5092870515152,
248
+ "per_object_best_kinematic_grasp_stability": {
249
+ "obj_014_lemon_2": 0.6950187650402412,
250
+ "obj_014_lemon_3": 0.5906668631727159,
251
+ "obj_017_orange_4": 0.3875350275843968,
252
+ "obj_017_orange_5": 0.44502823397944585
253
+ },
254
+ "grasp_gsi_diagnostics": {
255
+ "mean_anti_slip_score": 0.7113597543539685,
256
+ "mean_duration_score": 0.4175309791705706,
257
+ "mean_hold_score": 0.38286687504285577,
258
+ "mean_kinematic_stability_score": 0.14354139278313954,
259
+ "mean_legacy_gsi": 0.4553408432516696,
260
+ "mean_lift_score": 0.4573595670745583,
261
+ "mean_motion_stability_score": 0.37334461004458164
262
+ }
263
+ },
264
+ "tool": {
265
+ "tool_selection_accuracy": null,
266
+ "tool_switch_success_rate": null,
267
+ "tool_switch_total": 0,
268
+ "tool_switch_successful": 0
269
+ },
270
+ "runtime": {
271
+ "avg_steps": 1481.96,
272
+ "error_count": 0
273
+ }
274
+ }
275
+ }
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Equi/summary.json ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100800000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 46,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3432,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.92,
50
+ "success_rate_ci": {
51
+ "mean": 0.92,
52
+ "ci_low": 0.84,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.96
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 28.878623188405438
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.76,
64
+ "hard_violation_rate": 0.18,
65
+ "drop_rate": 0.18,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 9.632570933182067e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.92,
74
+ "at_end_success_rate": 0.92,
75
+ "task_family_success_rate": {
76
+ "pour_tilt": 0.92
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.0,
81
+ "current_chain_depth_progress_score": 0.0,
82
+ "latched_chain_depth_progress_score": 1.0,
83
+ "avg_current_chain_depth": 0,
84
+ "avg_latched_chain_depth": 1
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.323088611304915,
88
+ "expert_time_s": 38.13333333333333,
89
+ "expert_time_step": 2288,
90
+ "avg_steps_to_success": 1732.7173913043478,
91
+ "avg_policy_steps_to_success": 577.8913043478261,
92
+ "avg_policy_queries_to_success": 36.45652173913044
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 9.632570933182067e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_172_soy_sauce_1": 9
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.96,
101
+ "joint_limit_active_step_rate": 0.033221666862885704,
102
+ "finger_joint_limit_saturation_step_rate": 0.010670748022647244,
103
+ "max_joint_limit_excess_rad": 1.3252978324890137,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "left_hand_Thumb_Flexion": 100,
107
+ "left_hand_j3": 195,
108
+ "right_hand_Index_Finger_Distal": 581,
109
+ "right_hand_Index_Finger_Proximal": 232,
110
+ "right_hand_Middle_Finger_Distal": 638,
111
+ "right_hand_Middle_Finger_Proximal": 843,
112
+ "right_hand_Pinky": 516,
113
+ "right_hand_Ring_Finger": 1389,
114
+ "right_hand_Thumb_Flexion": 1513,
115
+ "right_hand_j12": 380,
116
+ "right_hand_j13": 326,
117
+ "right_hand_j3": 289
118
+ },
119
+ "joint_limit_hard_counts": {
120
+ "left_hand_Thumb_Flexion": 12,
121
+ "left_hand_Thumb_Opposition": 1,
122
+ "left_hand_j3": 9,
123
+ "left_hand_j5": 1,
124
+ "right_hand_Index_Finger_Distal": 81,
125
+ "right_hand_Index_Finger_Proximal": 43,
126
+ "right_hand_Middle_Finger_Distal": 95,
127
+ "right_hand_Middle_Finger_Proximal": 158,
128
+ "right_hand_Pinky": 52,
129
+ "right_hand_Ring_Finger": 132,
130
+ "right_hand_Thumb_Flexion": 129,
131
+ "right_hand_j12": 68,
132
+ "right_hand_j13": 50,
133
+ "right_hand_j3": 55
134
+ },
135
+ "joint_limit_hard_episode_events": 344,
136
+ "finger_joint_limit_saturation_counts": {
137
+ "left_hand_Thumb_Flexion": 24,
138
+ "left_hand_j3": 58,
139
+ "right_hand_Middle_Finger_Proximal": 401,
140
+ "right_hand_Pinky": 65,
141
+ "right_hand_Ring_Finger": 878,
142
+ "right_hand_Thumb_Flexion": 30
143
+ },
144
+ "joint_limit_max_excess_by_joint": {
145
+ "left_hand_Finger_Spread": 0.0025826720520853996,
146
+ "left_hand_Index_Finger_Distal": 0.0002221323666162789,
147
+ "left_hand_Index_Finger_Proximal": 0.0013137459754943848,
148
+ "left_hand_Middle_Finger_Proximal": 0.0011134147644042969,
149
+ "left_hand_Thumb_Flexion": 0.5490937829017639,
150
+ "left_hand_Thumb_Opposition": 0.00517757935449481,
151
+ "left_hand_index_spread": 0.004812332335859537,
152
+ "left_hand_j14": 1.9123417587252334e-05,
153
+ "left_hand_j3": 0.973770022392273,
154
+ "left_hand_j4": 8.134515283586552e-09,
155
+ "left_hand_j5": 0.005921096075326204,
156
+ "left_hand_ring_spread": 0.004414531867951155,
157
+ "right_hand_Finger_Spread": 0.0008874657796695828,
158
+ "right_hand_Index_Finger_Distal": 0.2071668952703476,
159
+ "right_hand_Index_Finger_Proximal": 0.15604911744594574,
160
+ "right_hand_Middle_Finger_Distal": 0.24186985194683075,
161
+ "right_hand_Middle_Finger_Proximal": 0.3833141028881073,
162
+ "right_hand_Pinky": 0.348781019449234,
163
+ "right_hand_Ring_Finger": 1.3252978324890137,
164
+ "right_hand_Thumb_Flexion": 0.10629560798406601,
165
+ "right_hand_index_spread": 0.0011255404679104686,
166
+ "right_hand_j12": 0.1338297426700592,
167
+ "right_hand_j13": 0.16750051081180573,
168
+ "right_hand_j14": 9.107702499022707e-05,
169
+ "right_hand_j15": 0.0001432000135537237,
170
+ "right_hand_j16": 6.91905224812217e-05,
171
+ "right_hand_j17": 5.1900533435400575e-05,
172
+ "right_hand_j3": 0.10829825699329376,
173
+ "right_hand_ring_spread": 0.002875754376873374
174
+ },
175
+ "coverage": {
176
+ "hard_limit_source": {
177
+ "available": 50,
178
+ "unavailable": 0,
179
+ "unknown": 0
180
+ },
181
+ "joint_role_manifest": {
182
+ "available": 50,
183
+ "unavailable": 0,
184
+ "unknown": 0
185
+ }
186
+ },
187
+ "hard_limit_sources": {
188
+ "runtime.joint_pos_limits": 50
189
+ },
190
+ "soft_limit_sources": {
191
+ "runtime.soft_joint_pos_limits": 50
192
+ },
193
+ "unknown_runtime_joint_names": [],
194
+ "baseline_checks": {
195
+ "post_home": {
196
+ "checked": 50,
197
+ "hard_violation_episodes": 0
198
+ },
199
+ "post_reset": {
200
+ "checked": 50,
201
+ "hard_violation_episodes": 0
202
+ }
203
+ }
204
+ }
205
+ },
206
+ "robot_motion": {
207
+ "joint_acceleration_rms": 12.062436202628215,
208
+ "joint_effort_rms": 23.175574445327964,
209
+ "joint_jerk_rms": 1174.19386094154,
210
+ "joint_velocity_rms": 0.3359369757627425
211
+ },
212
+ "grasp": {
213
+ "kinematic_grasp_stability_index": 0.9134922379282087,
214
+ "mean_kinematic_grasp_stability": 0.4328848363738145,
215
+ "per_object_best_kinematic_grasp_stability": {
216
+ "obj_146_olive_oil_1": 0.867616363505255,
217
+ "obj_172_soy_sauce_1": 0.8579695000402512
218
+ },
219
+ "grasp_gsi_diagnostics": {
220
+ "mean_anti_slip_score": 0.9889492931964277,
221
+ "mean_duration_score": 0.364852832941657,
222
+ "mean_hold_score": 0.11837196621003891,
223
+ "mean_kinematic_stability_score": 0.0764237068560624,
224
+ "mean_legacy_gsi": 0.339414429873879,
225
+ "mean_lift_score": 0.16985051147574534,
226
+ "mean_motion_stability_score": 0.1035473192424399
227
+ }
228
+ },
229
+ "tool": {
230
+ "tool_selection_accuracy": 1.0,
231
+ "tool_switch_success_rate": 0.0,
232
+ "tool_switch_total": 248,
233
+ "tool_switch_successful": 0
234
+ },
235
+ "runtime": {
236
+ "avg_steps": 1868.66,
237
+ "error_count": 0
238
+ }
239
+ }
240
+ }
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Full/summary.json ADDED
@@ -0,0 +1,241 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 200800000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 46,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3432,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.92,
50
+ "success_rate_ci": {
51
+ "mean": 0.92,
52
+ "ci_low": 0.84,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.95
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 29.102898550724266
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.84,
64
+ "hard_violation_rate": 0.12,
65
+ "drop_rate": 0.12,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 6.379449666142135e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.92,
74
+ "at_end_success_rate": 0.92,
75
+ "task_family_success_rate": {
76
+ "pour_tilt": 0.92
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.0,
81
+ "current_chain_depth_progress_score": 0.0,
82
+ "latched_chain_depth_progress_score": 0.98,
83
+ "avg_current_chain_depth": 0,
84
+ "avg_latched_chain_depth": 0.98
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.3159627769567357,
88
+ "expert_time_s": 38.13333333333333,
89
+ "expert_time_step": 2288,
90
+ "avg_steps_to_success": 1746.1739130434783,
91
+ "avg_policy_steps_to_success": 582.3478260869565,
92
+ "avg_policy_queries_to_success": 36.82608695652174
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 6.379449666142135e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_146_olive_oil_1": 2,
98
+ "drop_to_table_obj_172_soy_sauce_1": 4
99
+ },
100
+ "robot_constraint_violation_rate": 1.0,
101
+ "finger_joint_limit_saturation_rate": 1.0,
102
+ "joint_limit_active_step_rate": 0.024358865308552715,
103
+ "finger_joint_limit_saturation_step_rate": 0.00858035980096117,
104
+ "max_joint_limit_excess_rad": 0.5294941663742065,
105
+ "robot_constraint_diagnostics": {
106
+ "joint_limit_soft_counts": {
107
+ "left_hand_Thumb_Flexion": 160,
108
+ "left_hand_j3": 101,
109
+ "right_hand_Index_Finger_Distal": 594,
110
+ "right_hand_Index_Finger_Proximal": 251,
111
+ "right_hand_Middle_Finger_Distal": 668,
112
+ "right_hand_Middle_Finger_Proximal": 867,
113
+ "right_hand_Pinky": 427,
114
+ "right_hand_Ring_Finger": 1190,
115
+ "right_hand_Thumb_Flexion": 989,
116
+ "right_hand_j12": 420,
117
+ "right_hand_j13": 355,
118
+ "right_hand_j3": 316
119
+ },
120
+ "joint_limit_hard_counts": {
121
+ "left_hand_Thumb_Flexion": 9,
122
+ "left_hand_Thumb_Opposition": 10,
123
+ "left_hand_index_spread": 2,
124
+ "left_hand_j3": 7,
125
+ "left_hand_j5": 3,
126
+ "right_hand_Index_Finger_Distal": 80,
127
+ "right_hand_Index_Finger_Proximal": 48,
128
+ "right_hand_Middle_Finger_Distal": 94,
129
+ "right_hand_Middle_Finger_Proximal": 173,
130
+ "right_hand_Pinky": 53,
131
+ "right_hand_Ring_Finger": 126,
132
+ "right_hand_Thumb_Flexion": 116,
133
+ "right_hand_j12": 67,
134
+ "right_hand_j13": 52,
135
+ "right_hand_j3": 51
136
+ },
137
+ "joint_limit_hard_episode_events": 336,
138
+ "finger_joint_limit_saturation_counts": {
139
+ "left_hand_Thumb_Flexion": 32,
140
+ "left_hand_j3": 27,
141
+ "right_hand_Middle_Finger_Proximal": 410,
142
+ "right_hand_Ring_Finger": 727,
143
+ "right_hand_Thumb_Flexion": 17
144
+ },
145
+ "joint_limit_max_excess_by_joint": {
146
+ "left_hand_Finger_Spread": 0.0007727881311438978,
147
+ "left_hand_Index_Finger_Distal": 0.00015181985509116203,
148
+ "left_hand_Index_Finger_Proximal": 0.0015292763710021973,
149
+ "left_hand_Middle_Finger_Proximal": 0.0006070137023925781,
150
+ "left_hand_Thumb_Flexion": 0.1922380030155182,
151
+ "left_hand_Thumb_Opposition": 0.009061919525265694,
152
+ "left_hand_index_spread": 0.005548452492803335,
153
+ "left_hand_j14": 9.521131687506568e-06,
154
+ "left_hand_j3": 0.3356837332248688,
155
+ "left_hand_j4": 8.22512546960752e-09,
156
+ "left_hand_j5": 0.006459713447839022,
157
+ "left_hand_ring_spread": 0.004322588909417391,
158
+ "right_hand_Finger_Spread": 0.0007964828982949257,
159
+ "right_hand_Index_Finger_Distal": 0.2025892585515976,
160
+ "right_hand_Index_Finger_Proximal": 0.16924606263637543,
161
+ "right_hand_Middle_Finger_Distal": 0.23367621004581451,
162
+ "right_hand_Middle_Finger_Proximal": 0.40663671493530273,
163
+ "right_hand_Pinky": 0.17084647715091705,
164
+ "right_hand_Ring_Finger": 0.5294941663742065,
165
+ "right_hand_Thumb_Flexion": 0.1840752810239792,
166
+ "right_hand_index_spread": 0.0012347494484856725,
167
+ "right_hand_j12": 0.12590399384498596,
168
+ "right_hand_j13": 0.149588480591774,
169
+ "right_hand_j14": 0.004484417382627726,
170
+ "right_hand_j15": 0.00014663461479358375,
171
+ "right_hand_j16": 9.577524906489998e-05,
172
+ "right_hand_j17": 2.408497493888717e-05,
173
+ "right_hand_j3": 0.10440495610237122,
174
+ "right_hand_ring_spread": 0.0031649372540414333
175
+ },
176
+ "coverage": {
177
+ "hard_limit_source": {
178
+ "available": 50,
179
+ "unavailable": 0,
180
+ "unknown": 0
181
+ },
182
+ "joint_role_manifest": {
183
+ "available": 50,
184
+ "unavailable": 0,
185
+ "unknown": 0
186
+ }
187
+ },
188
+ "hard_limit_sources": {
189
+ "runtime.joint_pos_limits": 50
190
+ },
191
+ "soft_limit_sources": {
192
+ "runtime.soft_joint_pos_limits": 50
193
+ },
194
+ "unknown_runtime_joint_names": [],
195
+ "baseline_checks": {
196
+ "post_home": {
197
+ "checked": 50,
198
+ "hard_violation_episodes": 0
199
+ },
200
+ "post_reset": {
201
+ "checked": 50,
202
+ "hard_violation_episodes": 0
203
+ }
204
+ }
205
+ }
206
+ },
207
+ "robot_motion": {
208
+ "joint_acceleration_rms": 12.12375823143615,
209
+ "joint_effort_rms": 23.076521146917493,
210
+ "joint_jerk_rms": 1183.9680403644732,
211
+ "joint_velocity_rms": 0.33632170499966607
212
+ },
213
+ "grasp": {
214
+ "kinematic_grasp_stability_index": 0.9125874437369885,
215
+ "mean_kinematic_grasp_stability": 0.39991496103607344,
216
+ "per_object_best_kinematic_grasp_stability": {
217
+ "obj_146_olive_oil_1": 0.8800790327015191,
218
+ "obj_172_soy_sauce_1": 0.8300615635644658
219
+ },
220
+ "grasp_gsi_diagnostics": {
221
+ "mean_anti_slip_score": 0.9722878815136546,
222
+ "mean_duration_score": 0.36157316344985807,
223
+ "mean_hold_score": 0.10814644832764521,
224
+ "mean_kinematic_stability_score": 0.06501648885331167,
225
+ "mean_legacy_gsi": 0.3302323663536617,
226
+ "mean_lift_score": 0.1624339367589757,
227
+ "mean_motion_stability_score": 0.09544609738466149
228
+ }
229
+ },
230
+ "tool": {
231
+ "tool_selection_accuracy": 1.0,
232
+ "tool_switch_success_rate": 0.0,
233
+ "tool_switch_total": 248,
234
+ "tool_switch_successful": 0
235
+ },
236
+ "runtime": {
237
+ "avg_steps": 1881.04,
238
+ "error_count": 0
239
+ }
240
+ }
241
+ }
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/08_frypan_stand_pour/Inv/summary.json ADDED
@@ -0,0 +1,239 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100800000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 48,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3432,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.96,
50
+ "success_rate_ci": {
51
+ "mean": 0.96,
52
+ "ci_low": 0.9,
53
+ "ci_high": 1.0
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.97
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 28.842361111110755
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.9,
64
+ "hard_violation_rate": 0.08,
65
+ "drop_rate": 0.08,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 4.4479039252752143e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.96,
74
+ "at_end_success_rate": 0.96,
75
+ "task_family_success_rate": {
76
+ "pour_tilt": 0.96
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.0,
81
+ "current_chain_depth_progress_score": 0.0,
82
+ "latched_chain_depth_progress_score": 0.98,
83
+ "avg_current_chain_depth": 0,
84
+ "avg_latched_chain_depth": 0.98
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.324624469313869,
88
+ "expert_time_s": 38.13333333333333,
89
+ "expert_time_step": 2288,
90
+ "avg_steps_to_success": 1730.5416666666667,
91
+ "avg_policy_steps_to_success": 577.1666666666666,
92
+ "avg_policy_queries_to_success": 36.416666666666664
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 4.4479039252752143e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_172_soy_sauce_1": 4
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.96,
101
+ "joint_limit_active_step_rate": 0.023951962637607026,
102
+ "finger_joint_limit_saturation_step_rate": 0.0076726342710997444,
103
+ "max_joint_limit_excess_rad": 0.5221675634384155,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "left_hand_Thumb_Flexion": 121,
107
+ "left_hand_j3": 157,
108
+ "right_hand_Index_Finger_Distal": 588,
109
+ "right_hand_Index_Finger_Proximal": 229,
110
+ "right_hand_Middle_Finger_Distal": 644,
111
+ "right_hand_Middle_Finger_Proximal": 837,
112
+ "right_hand_Pinky": 391,
113
+ "right_hand_Ring_Finger": 1138,
114
+ "right_hand_Thumb_Flexion": 901,
115
+ "right_hand_j12": 388,
116
+ "right_hand_j13": 324,
117
+ "right_hand_j3": 290
118
+ },
119
+ "joint_limit_hard_counts": {
120
+ "left_hand_Thumb_Flexion": 5,
121
+ "left_hand_Thumb_Opposition": 4,
122
+ "left_hand_index_spread": 1,
123
+ "left_hand_j3": 8,
124
+ "left_hand_j5": 4,
125
+ "right_hand_Index_Finger_Distal": 82,
126
+ "right_hand_Index_Finger_Proximal": 37,
127
+ "right_hand_Middle_Finger_Distal": 92,
128
+ "right_hand_Middle_Finger_Proximal": 168,
129
+ "right_hand_Pinky": 49,
130
+ "right_hand_Ring_Finger": 118,
131
+ "right_hand_Thumb_Flexion": 138,
132
+ "right_hand_j12": 63,
133
+ "right_hand_j13": 47,
134
+ "right_hand_j3": 51
135
+ },
136
+ "joint_limit_hard_episode_events": 336,
137
+ "finger_joint_limit_saturation_counts": {
138
+ "left_hand_Thumb_Flexion": 2,
139
+ "left_hand_j3": 31,
140
+ "right_hand_Middle_Finger_Proximal": 396,
141
+ "right_hand_Ring_Finger": 655
142
+ },
143
+ "joint_limit_max_excess_by_joint": {
144
+ "left_hand_Finger_Spread": 0.0005932403728365898,
145
+ "left_hand_Index_Finger_Distal": 0.00015866561443544924,
146
+ "left_hand_Index_Finger_Proximal": 0.0009046196937561035,
147
+ "left_hand_Middle_Finger_Proximal": 0.0009998679161071777,
148
+ "left_hand_Thumb_Flexion": 0.11833306401968002,
149
+ "left_hand_Thumb_Opposition": 0.00788294430822134,
150
+ "left_hand_index_spread": 0.005929997190833092,
151
+ "left_hand_j14": 4.232876017340459e-05,
152
+ "left_hand_j3": 0.2917395234107971,
153
+ "left_hand_j4": 9.33640187383844e-09,
154
+ "left_hand_j5": 0.005781437736004591,
155
+ "left_hand_ring_spread": 0.004097987897694111,
156
+ "right_hand_Finger_Spread": 0.0013159919762983918,
157
+ "right_hand_Index_Finger_Distal": 0.1875271201133728,
158
+ "right_hand_Index_Finger_Proximal": 0.13421331346035004,
159
+ "right_hand_Middle_Finger_Distal": 0.23442666232585907,
160
+ "right_hand_Middle_Finger_Proximal": 0.36812999844551086,
161
+ "right_hand_Pinky": 0.1586095690727234,
162
+ "right_hand_Ring_Finger": 0.5221675634384155,
163
+ "right_hand_Thumb_Flexion": 0.10386160016059875,
164
+ "right_hand_index_spread": 0.0011855217162519693,
165
+ "right_hand_j12": 0.13349084556102753,
166
+ "right_hand_j13": 0.1418139487504959,
167
+ "right_hand_j14": 8.99869337445125e-05,
168
+ "right_hand_j15": 0.0001477327023167163,
169
+ "right_hand_j16": 0.0001011398562695831,
170
+ "right_hand_j17": 7.38506278139539e-05,
171
+ "right_hand_j3": 0.108154296875,
172
+ "right_hand_ring_spread": 0.002014011377468705
173
+ },
174
+ "coverage": {
175
+ "hard_limit_source": {
176
+ "available": 50,
177
+ "unavailable": 0,
178
+ "unknown": 0
179
+ },
180
+ "joint_role_manifest": {
181
+ "available": 50,
182
+ "unavailable": 0,
183
+ "unknown": 0
184
+ }
185
+ },
186
+ "hard_limit_sources": {
187
+ "runtime.joint_pos_limits": 50
188
+ },
189
+ "soft_limit_sources": {
190
+ "runtime.soft_joint_pos_limits": 50
191
+ },
192
+ "unknown_runtime_joint_names": [],
193
+ "baseline_checks": {
194
+ "post_home": {
195
+ "checked": 50,
196
+ "hard_violation_episodes": 0
197
+ },
198
+ "post_reset": {
199
+ "checked": 50,
200
+ "hard_violation_episodes": 0
201
+ }
202
+ }
203
+ }
204
+ },
205
+ "robot_motion": {
206
+ "joint_acceleration_rms": 12.135664248932374,
207
+ "joint_effort_rms": 23.06345703835884,
208
+ "joint_jerk_rms": 1183.9578182376467,
209
+ "joint_velocity_rms": 0.33447849418227055
210
+ },
211
+ "grasp": {
212
+ "kinematic_grasp_stability_index": 0.9107072016644215,
213
+ "mean_kinematic_grasp_stability": 0.4093359460725043,
214
+ "per_object_best_kinematic_grasp_stability": {
215
+ "obj_146_olive_oil_1": 0.8884434619090028,
216
+ "obj_172_soy_sauce_1": 0.8487974445512009
217
+ },
218
+ "grasp_gsi_diagnostics": {
219
+ "mean_anti_slip_score": 0.978147354325359,
220
+ "mean_duration_score": 0.3568072424978502,
221
+ "mean_hold_score": 0.11504614459439429,
222
+ "mean_kinematic_stability_score": 0.06921512019639099,
223
+ "mean_legacy_gsi": 0.3342282139777003,
224
+ "mean_lift_score": 0.16852873613225067,
225
+ "mean_motion_stability_score": 0.1014655536073534
226
+ }
227
+ },
228
+ "tool": {
229
+ "tool_selection_accuracy": 1.0,
230
+ "tool_switch_success_rate": 0.0,
231
+ "tool_switch_total": 249,
232
+ "tool_switch_successful": 0
233
+ },
234
+ "runtime": {
235
+ "avg_steps": 1798.6,
236
+ "error_count": 0
237
+ }
238
+ }
239
+ }
me-dex-1-5/per_task_eval/08_frypan_stand_pour/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/08_frypan_stand_pour/None/summary.json ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100800000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 48,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3432,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.96,
50
+ "success_rate_ci": {
51
+ "mean": 0.96,
52
+ "ci_low": 0.9,
53
+ "ci_high": 1.0
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.97
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 29.303124999999618
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.82,
64
+ "hard_violation_rate": 0.16,
65
+ "drop_rate": 0.16,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 8.766450792815893e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.96,
74
+ "at_end_success_rate": 0.96,
75
+ "task_family_success_rate": {
76
+ "pour_tilt": 0.96
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.0,
81
+ "current_chain_depth_progress_score": 0.0,
82
+ "latched_chain_depth_progress_score": 0.98,
83
+ "avg_current_chain_depth": 0,
84
+ "avg_latched_chain_depth": 0.98
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 1.304941369343521,
88
+ "expert_time_s": 38.13333333333333,
89
+ "expert_time_step": 2288,
90
+ "avg_steps_to_success": 1758.1875,
91
+ "avg_policy_steps_to_success": 586.4166666666666,
92
+ "avg_policy_queries_to_success": 37.0625
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 8.766450792815893e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_146_olive_oil_1": 1,
98
+ "drop_to_table_obj_172_soy_sauce_1": 7
99
+ },
100
+ "robot_constraint_violation_rate": 1.0,
101
+ "finger_joint_limit_saturation_rate": 0.96,
102
+ "joint_limit_active_step_rate": 0.02639797494986686,
103
+ "finger_joint_limit_saturation_step_rate": 0.008470583078558358,
104
+ "max_joint_limit_excess_rad": 0.5325779914855957,
105
+ "robot_constraint_diagnostics": {
106
+ "joint_limit_soft_counts": {
107
+ "left_hand_Thumb_Flexion": 38,
108
+ "left_hand_j3": 121,
109
+ "right_hand_Index_Finger_Distal": 568,
110
+ "right_hand_Index_Finger_Proximal": 209,
111
+ "right_hand_Middle_Finger_Distal": 629,
112
+ "right_hand_Middle_Finger_Proximal": 853,
113
+ "right_hand_Pinky": 475,
114
+ "right_hand_Ring_Finger": 1083,
115
+ "right_hand_Thumb_Flexion": 1204,
116
+ "right_hand_j12": 356,
117
+ "right_hand_j13": 306,
118
+ "right_hand_j3": 276
119
+ },
120
+ "joint_limit_hard_counts": {
121
+ "left_hand_Thumb_Flexion": 6,
122
+ "left_hand_Thumb_Opposition": 2,
123
+ "left_hand_j3": 5,
124
+ "left_hand_j5": 2,
125
+ "right_hand_Index_Finger_Distal": 71,
126
+ "right_hand_Index_Finger_Proximal": 40,
127
+ "right_hand_Middle_Finger_Distal": 96,
128
+ "right_hand_Middle_Finger_Proximal": 173,
129
+ "right_hand_Pinky": 55,
130
+ "right_hand_Ring_Finger": 122,
131
+ "right_hand_Thumb_Flexion": 91,
132
+ "right_hand_j12": 66,
133
+ "right_hand_j13": 50,
134
+ "right_hand_j3": 55
135
+ },
136
+ "joint_limit_hard_episode_events": 292,
137
+ "finger_joint_limit_saturation_counts": {
138
+ "left_hand_j3": 55,
139
+ "right_hand_Middle_Finger_Proximal": 385,
140
+ "right_hand_Pinky": 23,
141
+ "right_hand_Ring_Finger": 664,
142
+ "right_hand_Thumb_Flexion": 31
143
+ },
144
+ "joint_limit_max_excess_by_joint": {
145
+ "left_hand_Finger_Spread": 0.000717418675776571,
146
+ "left_hand_Index_Finger_Distal": 0.00012954972044099122,
147
+ "left_hand_Index_Finger_Proximal": 0.0006485581398010254,
148
+ "left_hand_Middle_Finger_Proximal": 0.0012157559394836426,
149
+ "left_hand_Thumb_Flexion": 0.057231880724430084,
150
+ "left_hand_Thumb_Opposition": 0.006378688849508762,
151
+ "left_hand_index_spread": 0.004957862198352814,
152
+ "left_hand_j14": 3.8592010241700336e-05,
153
+ "left_hand_j3": 0.48276486992836,
154
+ "left_hand_j4": 9.329577999039884e-09,
155
+ "left_hand_j5": 0.0058939713053405285,
156
+ "left_hand_ring_spread": 0.0032048013526946306,
157
+ "right_hand_Finger_Spread": 0.0010853075655177236,
158
+ "right_hand_Index_Finger_Distal": 0.19636908173561096,
159
+ "right_hand_Index_Finger_Proximal": 0.15461009740829468,
160
+ "right_hand_Middle_Finger_Distal": 0.23373103141784668,
161
+ "right_hand_Middle_Finger_Proximal": 0.41339755058288574,
162
+ "right_hand_Pinky": 0.2544288635253906,
163
+ "right_hand_Ring_Finger": 0.5325779914855957,
164
+ "right_hand_Thumb_Flexion": 0.1458963006734848,
165
+ "right_hand_index_spread": 0.0011173341190442443,
166
+ "right_hand_j12": 0.12676121294498444,
167
+ "right_hand_j13": 0.1377885937690735,
168
+ "right_hand_j14": 0.00022581145458389074,
169
+ "right_hand_j15": 0.0001520298101240769,
170
+ "right_hand_j16": 5.978295666864142e-05,
171
+ "right_hand_j17": 6.640223728027195e-05,
172
+ "right_hand_j3": 0.17288340628147125,
173
+ "right_hand_ring_spread": 0.002647604327648878
174
+ },
175
+ "coverage": {
176
+ "hard_limit_source": {
177
+ "available": 50,
178
+ "unavailable": 0,
179
+ "unknown": 0
180
+ },
181
+ "joint_role_manifest": {
182
+ "available": 50,
183
+ "unavailable": 0,
184
+ "unknown": 0
185
+ }
186
+ },
187
+ "hard_limit_sources": {
188
+ "runtime.joint_pos_limits": 50
189
+ },
190
+ "soft_limit_sources": {
191
+ "runtime.soft_joint_pos_limits": 50
192
+ },
193
+ "unknown_runtime_joint_names": [],
194
+ "baseline_checks": {
195
+ "post_home": {
196
+ "checked": 50,
197
+ "hard_violation_episodes": 0
198
+ },
199
+ "post_reset": {
200
+ "checked": 50,
201
+ "hard_violation_episodes": 0
202
+ }
203
+ }
204
+ }
205
+ },
206
+ "robot_motion": {
207
+ "joint_acceleration_rms": 11.929962218680586,
208
+ "joint_effort_rms": 23.12017167832075,
209
+ "joint_jerk_rms": 1161.8316903983498,
210
+ "joint_velocity_rms": 0.3421571650200721
211
+ },
212
+ "grasp": {
213
+ "kinematic_grasp_stability_index": 0.9157285515353429,
214
+ "mean_kinematic_grasp_stability": 0.4404322940879842,
215
+ "per_object_best_kinematic_grasp_stability": {
216
+ "obj_146_olive_oil_1": 0.9021810385182338,
217
+ "obj_172_soy_sauce_1": 0.8400631060566236
218
+ },
219
+ "grasp_gsi_diagnostics": {
220
+ "mean_anti_slip_score": 0.9832962906565359,
221
+ "mean_duration_score": 0.35981546722010405,
222
+ "mean_hold_score": 0.12147629277639385,
223
+ "mean_kinematic_stability_score": 0.07766208813581875,
224
+ "mean_legacy_gsi": 0.33814223507653696,
225
+ "mean_lift_score": 0.17098625650908628,
226
+ "mean_motion_stability_score": 0.10408536927587997
227
+ }
228
+ },
229
+ "tool": {
230
+ "tool_selection_accuracy": 1.0,
231
+ "tool_switch_success_rate": 0.0,
232
+ "tool_switch_total": 252,
233
+ "tool_switch_successful": 0
234
+ },
235
+ "runtime": {
236
+ "avg_steps": 1825.14,
237
+ "error_count": 0
238
+ }
239
+ }
240
+ }
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Equi/summary.json ADDED
@@ -0,0 +1,229 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100900000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 35,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1926,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.7,
50
+ "success_rate_ci": {
51
+ "mean": 0.7,
52
+ "ci_low": 0.5795000000000005,
53
+ "ci_high": 0.82
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.95
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.532380952380898
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.7,
64
+ "hard_violation_rate": 0.02,
65
+ "drop_rate": 0.02,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 1.2770087347397456e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.72,
74
+ "at_end_success_rate": 0.7,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.7
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.82,
81
+ "current_chain_depth_progress_score": 0.9,
82
+ "latched_chain_depth_progress_score": 0.95,
83
+ "avg_current_chain_depth": 1.8,
84
+ "avg_latched_chain_depth": 1.9
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9144123413016867,
88
+ "expert_time_s": 21.383333333333333,
89
+ "expert_time_step": 1283,
90
+ "avg_steps_to_success": 1411.942857142857,
91
+ "avg_policy_steps_to_success": 471,
92
+ "avg_policy_queries_to_success": 29.82857142857143
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 1.2770087347397456e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_209_soap_3": 1
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.96,
101
+ "joint_limit_active_step_rate": 0.10274812279715993,
102
+ "finger_joint_limit_saturation_step_rate": 0.05858916074985953,
103
+ "max_joint_limit_excess_rad": 1.7330644130706787,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "leap_l_10": 1414,
107
+ "leap_l_2": 288,
108
+ "multi_leap_r_1": 829,
109
+ "multi_leap_r_10": 3433,
110
+ "multi_leap_r_12": 95,
111
+ "multi_leap_r_13": 709,
112
+ "multi_leap_r_14": 15,
113
+ "multi_leap_r_2": 2225,
114
+ "multi_leap_r_5": 1065,
115
+ "multi_leap_r_6": 242,
116
+ "multi_leap_r_9": 2070
117
+ },
118
+ "joint_limit_hard_counts": {
119
+ "leap_l_10": 61,
120
+ "leap_l_2": 288,
121
+ "multi_leap_r_1": 23,
122
+ "multi_leap_r_10": 225,
123
+ "multi_leap_r_12": 2,
124
+ "multi_leap_r_13": 14,
125
+ "multi_leap_r_14": 1,
126
+ "multi_leap_r_2": 93,
127
+ "multi_leap_r_3": 1,
128
+ "multi_leap_r_5": 30,
129
+ "multi_leap_r_6": 5,
130
+ "multi_leap_r_9": 41
131
+ },
132
+ "joint_limit_hard_episode_events": 610,
133
+ "finger_joint_limit_saturation_counts": {
134
+ "leap_l_10": 849,
135
+ "multi_leap_r_1": 151,
136
+ "multi_leap_r_10": 2224,
137
+ "multi_leap_r_12": 70,
138
+ "multi_leap_r_13": 607,
139
+ "multi_leap_r_14": 3,
140
+ "multi_leap_r_2": 1402,
141
+ "multi_leap_r_5": 334,
142
+ "multi_leap_r_6": 75,
143
+ "multi_leap_r_9": 839
144
+ },
145
+ "joint_limit_max_excess_by_joint": {
146
+ "leap_l_10": 0.8761228322982788,
147
+ "leap_l_11": 0.0010790824890136719,
148
+ "leap_l_13": 0.004925727844238281,
149
+ "leap_l_15": 0.00030219554901123047,
150
+ "leap_l_2": 0.12071084976196289,
151
+ "leap_l_3": 3.987550735473633e-05,
152
+ "multi_leap_r_1": 0.3694652318954468,
153
+ "multi_leap_r_10": 0.9338403940200806,
154
+ "multi_leap_r_11": 0.004718393087387085,
155
+ "multi_leap_r_12": 1.7330644130706787,
156
+ "multi_leap_r_13": 0.9268143177032471,
157
+ "multi_leap_r_14": 0.25998806953430176,
158
+ "multi_leap_r_2": 0.8443387746810913,
159
+ "multi_leap_r_3": 0.008005529642105103,
160
+ "multi_leap_r_5": 0.3757205605506897,
161
+ "multi_leap_r_6": 0.34357815980911255,
162
+ "multi_leap_r_9": 0.47497570514678955
163
+ },
164
+ "coverage": {
165
+ "hard_limit_source": {
166
+ "available": 50,
167
+ "unavailable": 0,
168
+ "unknown": 0
169
+ },
170
+ "joint_role_manifest": {
171
+ "available": 50,
172
+ "unavailable": 0,
173
+ "unknown": 0
174
+ }
175
+ },
176
+ "hard_limit_sources": {
177
+ "runtime.joint_pos_limits": 50
178
+ },
179
+ "soft_limit_sources": {
180
+ "runtime.soft_joint_pos_limits": 50
181
+ },
182
+ "unknown_runtime_joint_names": [],
183
+ "baseline_checks": {
184
+ "post_home": {
185
+ "checked": 50,
186
+ "hard_violation_episodes": 0
187
+ },
188
+ "post_reset": {
189
+ "checked": 50,
190
+ "hard_violation_episodes": 0
191
+ }
192
+ }
193
+ }
194
+ },
195
+ "robot_motion": {
196
+ "joint_acceleration_rms": 18.70048067200572,
197
+ "joint_effort_rms": 64.52680715894698,
198
+ "joint_jerk_rms": 1845.6815163345248,
199
+ "joint_velocity_rms": 0.6841988514463473
200
+ },
201
+ "grasp": {
202
+ "kinematic_grasp_stability_index": 0.9726530436931351,
203
+ "mean_kinematic_grasp_stability": 0.5538422219839638,
204
+ "per_object_best_kinematic_grasp_stability": {
205
+ "obj_200_cleaner_2": 0.9064940847040007,
206
+ "obj_209_soap_3": 0.8909431666846366
207
+ },
208
+ "grasp_gsi_diagnostics": {
209
+ "mean_anti_slip_score": 0.7639518696622085,
210
+ "mean_duration_score": 0.45418087923842265,
211
+ "mean_hold_score": 0.4358836657555887,
212
+ "mean_kinematic_stability_score": 0.15851895760880158,
213
+ "mean_legacy_gsi": 0.49851886887388197,
214
+ "mean_lift_score": 0.4920556761191531,
215
+ "mean_motion_stability_score": 0.4034021720712085
216
+ }
217
+ },
218
+ "tool": {
219
+ "tool_selection_accuracy": null,
220
+ "tool_switch_success_rate": null,
221
+ "tool_switch_total": 0,
222
+ "tool_switch_successful": 0
223
+ },
224
+ "runtime": {
225
+ "avg_steps": 1566.16,
226
+ "error_count": 0
227
+ }
228
+ }
229
+ }
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Full/summary.json ADDED
@@ -0,0 +1,237 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 200900000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 31,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1926,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.62,
50
+ "success_rate_ci": {
51
+ "mean": 0.62,
52
+ "ci_low": 0.48,
53
+ "ci_high": 0.74
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.92
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.1327956989247
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.62,
64
+ "hard_violation_rate": 0.08,
65
+ "drop_rate": 0.08,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 5.023800253701913e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.62,
74
+ "at_end_success_rate": 0.62,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.62
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.77,
81
+ "current_chain_depth_progress_score": 0.86,
82
+ "latched_chain_depth_progress_score": 0.92,
83
+ "avg_current_chain_depth": 1.72,
84
+ "avg_latched_chain_depth": 1.84
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.926531654919593,
88
+ "expert_time_s": 21.383333333333333,
89
+ "expert_time_step": 1283,
90
+ "avg_steps_to_success": 1387.967741935484,
91
+ "avg_policy_steps_to_success": 462.96774193548384,
92
+ "avg_policy_queries_to_success": 29.483870967741936
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 5.023800253701913e-05,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_209_soap_3": 4
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 0.98,
101
+ "joint_limit_active_step_rate": 0.11231961417214052,
102
+ "finger_joint_limit_saturation_step_rate": 0.06852463546049409,
103
+ "max_joint_limit_excess_rad": 1.5320544242858887,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "leap_l_10": 1307,
107
+ "leap_l_11": 3,
108
+ "leap_l_2": 300,
109
+ "leap_l_6": 1,
110
+ "leap_l_9": 6,
111
+ "multi_leap_r_1": 410,
112
+ "multi_leap_r_10": 3867,
113
+ "multi_leap_r_12": 684,
114
+ "multi_leap_r_13": 1564,
115
+ "multi_leap_r_14": 17,
116
+ "multi_leap_r_2": 2999,
117
+ "multi_leap_r_5": 1167,
118
+ "multi_leap_r_6": 646,
119
+ "multi_leap_r_9": 1785
120
+ },
121
+ "joint_limit_hard_counts": {
122
+ "leap_l_10": 60,
123
+ "leap_l_11": 1,
124
+ "leap_l_2": 301,
125
+ "leap_l_6": 1,
126
+ "leap_l_9": 1,
127
+ "multi_leap_r_1": 13,
128
+ "multi_leap_r_10": 206,
129
+ "multi_leap_r_11": 2,
130
+ "multi_leap_r_12": 10,
131
+ "multi_leap_r_13": 21,
132
+ "multi_leap_r_14": 2,
133
+ "multi_leap_r_2": 128,
134
+ "multi_leap_r_3": 3,
135
+ "multi_leap_r_5": 40,
136
+ "multi_leap_r_6": 11,
137
+ "multi_leap_r_9": 46
138
+ },
139
+ "joint_limit_hard_episode_events": 581,
140
+ "finger_joint_limit_saturation_counts": {
141
+ "leap_l_10": 867,
142
+ "multi_leap_r_1": 103,
143
+ "multi_leap_r_10": 2375,
144
+ "multi_leap_r_12": 617,
145
+ "multi_leap_r_13": 1391,
146
+ "multi_leap_r_2": 1951,
147
+ "multi_leap_r_5": 201,
148
+ "multi_leap_r_6": 308,
149
+ "multi_leap_r_9": 564
150
+ },
151
+ "joint_limit_max_excess_by_joint": {
152
+ "leap_l_10": 0.7844966650009155,
153
+ "leap_l_11": 0.022901535034179688,
154
+ "leap_l_13": 0.0052187442779541016,
155
+ "leap_l_15": 0.000247955322265625,
156
+ "leap_l_2": 0.1811765432357788,
157
+ "leap_l_3": 3.552436828613281e-05,
158
+ "leap_l_6": 0.012517452239990234,
159
+ "leap_l_9": 0.029496312141418457,
160
+ "multi_leap_r_1": 0.3326282501220703,
161
+ "multi_leap_r_10": 0.907849907875061,
162
+ "multi_leap_r_11": 0.006763845682144165,
163
+ "multi_leap_r_12": 1.5320544242858887,
164
+ "multi_leap_r_13": 0.9318808317184448,
165
+ "multi_leap_r_14": 0.12053775787353516,
166
+ "multi_leap_r_2": 1.3152971267700195,
167
+ "multi_leap_r_3": 0.007252633571624756,
168
+ "multi_leap_r_5": 0.3301336169242859,
169
+ "multi_leap_r_6": 0.37600213289260864,
170
+ "multi_leap_r_9": 0.44115108251571655
171
+ },
172
+ "coverage": {
173
+ "hard_limit_source": {
174
+ "available": 50,
175
+ "unavailable": 0,
176
+ "unknown": 0
177
+ },
178
+ "joint_role_manifest": {
179
+ "available": 50,
180
+ "unavailable": 0,
181
+ "unknown": 0
182
+ }
183
+ },
184
+ "hard_limit_sources": {
185
+ "runtime.joint_pos_limits": 50
186
+ },
187
+ "soft_limit_sources": {
188
+ "runtime.soft_joint_pos_limits": 50
189
+ },
190
+ "unknown_runtime_joint_names": [],
191
+ "baseline_checks": {
192
+ "post_home": {
193
+ "checked": 50,
194
+ "hard_violation_episodes": 0
195
+ },
196
+ "post_reset": {
197
+ "checked": 50,
198
+ "hard_violation_episodes": 0
199
+ }
200
+ }
201
+ }
202
+ },
203
+ "robot_motion": {
204
+ "joint_acceleration_rms": 18.110740417341248,
205
+ "joint_effort_rms": 65.30201396731246,
206
+ "joint_jerk_rms": 1776.9120552615686,
207
+ "joint_velocity_rms": 0.6843099169849731
208
+ },
209
+ "grasp": {
210
+ "kinematic_grasp_stability_index": 0.9558362607108726,
211
+ "mean_kinematic_grasp_stability": 0.5172968912637066,
212
+ "per_object_best_kinematic_grasp_stability": {
213
+ "obj_200_cleaner_2": 0.8944803520340381,
214
+ "obj_209_soap_3": 0.8331845760339049
215
+ },
216
+ "grasp_gsi_diagnostics": {
217
+ "mean_anti_slip_score": 0.7108577599860477,
218
+ "mean_duration_score": 0.45103721304739547,
219
+ "mean_hold_score": 0.43152123284283095,
220
+ "mean_kinematic_stability_score": 0.1133430378349171,
221
+ "mean_legacy_gsi": 0.48551859029523636,
222
+ "mean_lift_score": 0.48475982642867993,
223
+ "mean_motion_stability_score": 0.40163742253002677
224
+ }
225
+ },
226
+ "tool": {
227
+ "tool_selection_accuracy": null,
228
+ "tool_switch_success_rate": null,
229
+ "tool_switch_total": 0,
230
+ "tool_switch_successful": 0
231
+ },
232
+ "runtime": {
233
+ "avg_steps": 1592.42,
234
+ "error_count": 0
235
+ }
236
+ }
237
+ }
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/Inv/summary.json ADDED
@@ -0,0 +1,238 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100900000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 34,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1926,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.68,
50
+ "success_rate_ci": {
51
+ "mean": 0.68,
52
+ "ci_low": 0.54,
53
+ "ci_high": 0.8
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.93
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.54166666666661
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.68,
64
+ "hard_violation_rate": 0.02,
65
+ "drop_rate": 0.0,
66
+ "high_speed_violation_rate": 0.02,
67
+ "safety_violation_step_rate": 1.2683755913801194e-05
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.68,
74
+ "at_end_success_rate": 0.68,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.68
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.82,
81
+ "current_chain_depth_progress_score": 0.91,
82
+ "latched_chain_depth_progress_score": 0.93,
83
+ "avg_current_chain_depth": 1.82,
84
+ "avg_latched_chain_depth": 1.86
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9110348420707753,
88
+ "expert_time_s": 21.383333333333333,
89
+ "expert_time_step": 1283,
90
+ "avg_steps_to_success": 1412.5,
91
+ "avg_policy_steps_to_success": 471.2647058823529,
92
+ "avg_policy_queries_to_success": 29.970588235294116
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 1.2683755913801194e-05,
96
+ "violation_counts": {
97
+ "high_speed_obj_209_soap_3": 1
98
+ },
99
+ "robot_constraint_violation_rate": 1.0,
100
+ "finger_joint_limit_saturation_rate": 1.0,
101
+ "joint_limit_active_step_rate": 0.1167793406983676,
102
+ "finger_joint_limit_saturation_step_rate": 0.07450438223766821,
103
+ "max_joint_limit_excess_rad": 2.854825258255005,
104
+ "robot_constraint_diagnostics": {
105
+ "joint_limit_soft_counts": {
106
+ "leap_l_10": 1897,
107
+ "leap_l_2": 332,
108
+ "leap_l_6": 90,
109
+ "leap_l_9": 32,
110
+ "multi_leap_r_1": 358,
111
+ "multi_leap_r_10": 3727,
112
+ "multi_leap_r_12": 422,
113
+ "multi_leap_r_13": 1062,
114
+ "multi_leap_r_14": 123,
115
+ "multi_leap_r_2": 2663,
116
+ "multi_leap_r_5": 1188,
117
+ "multi_leap_r_6": 538,
118
+ "multi_leap_r_9": 2530
119
+ },
120
+ "joint_limit_hard_counts": {
121
+ "leap_l_10": 56,
122
+ "leap_l_11": 1,
123
+ "leap_l_2": 323,
124
+ "leap_l_6": 2,
125
+ "leap_l_9": 1,
126
+ "multi_leap_r_1": 19,
127
+ "multi_leap_r_10": 185,
128
+ "multi_leap_r_11": 4,
129
+ "multi_leap_r_12": 11,
130
+ "multi_leap_r_13": 15,
131
+ "multi_leap_r_14": 3,
132
+ "multi_leap_r_2": 96,
133
+ "multi_leap_r_5": 34,
134
+ "multi_leap_r_6": 12,
135
+ "multi_leap_r_9": 50
136
+ },
137
+ "joint_limit_hard_episode_events": 578,
138
+ "finger_joint_limit_saturation_counts": {
139
+ "leap_l_10": 1376,
140
+ "leap_l_2": 10,
141
+ "leap_l_6": 45,
142
+ "multi_leap_r_1": 88,
143
+ "multi_leap_r_10": 2413,
144
+ "multi_leap_r_12": 316,
145
+ "multi_leap_r_13": 944,
146
+ "multi_leap_r_14": 82,
147
+ "multi_leap_r_2": 1708,
148
+ "multi_leap_r_5": 395,
149
+ "multi_leap_r_6": 238,
150
+ "multi_leap_r_9": 1115
151
+ },
152
+ "joint_limit_max_excess_by_joint": {
153
+ "leap_l_10": 0.738611102104187,
154
+ "leap_l_11": 0.008184343576431274,
155
+ "leap_l_13": 0.004596710205078125,
156
+ "leap_l_15": 0.0004216432571411133,
157
+ "leap_l_2": 0.25969213247299194,
158
+ "leap_l_3": 0.00010246038436889648,
159
+ "leap_l_6": 0.8435003757476807,
160
+ "leap_l_9": 0.07128667831420898,
161
+ "multi_leap_r_1": 0.38194793462753296,
162
+ "multi_leap_r_10": 0.7388008832931519,
163
+ "multi_leap_r_11": 0.00602269172668457,
164
+ "multi_leap_r_12": 2.854825258255005,
165
+ "multi_leap_r_13": 1.1139562129974365,
166
+ "multi_leap_r_14": 1.6344280242919922,
167
+ "multi_leap_r_2": 1.1486482620239258,
168
+ "multi_leap_r_3": 0.003487825393676758,
169
+ "multi_leap_r_5": 0.4373672604560852,
170
+ "multi_leap_r_6": 0.5032252073287964,
171
+ "multi_leap_r_9": 0.6036291718482971
172
+ },
173
+ "coverage": {
174
+ "hard_limit_source": {
175
+ "available": 50,
176
+ "unavailable": 0,
177
+ "unknown": 0
178
+ },
179
+ "joint_role_manifest": {
180
+ "available": 50,
181
+ "unavailable": 0,
182
+ "unknown": 0
183
+ }
184
+ },
185
+ "hard_limit_sources": {
186
+ "runtime.joint_pos_limits": 50
187
+ },
188
+ "soft_limit_sources": {
189
+ "runtime.soft_joint_pos_limits": 50
190
+ },
191
+ "unknown_runtime_joint_names": [],
192
+ "baseline_checks": {
193
+ "post_home": {
194
+ "checked": 50,
195
+ "hard_violation_episodes": 0
196
+ },
197
+ "post_reset": {
198
+ "checked": 50,
199
+ "hard_violation_episodes": 0
200
+ }
201
+ }
202
+ }
203
+ },
204
+ "robot_motion": {
205
+ "joint_acceleration_rms": 17.996705501331064,
206
+ "joint_effort_rms": 65.4737998626129,
207
+ "joint_jerk_rms": 1767.3502763737363,
208
+ "joint_velocity_rms": 0.6862296713615684
209
+ },
210
+ "grasp": {
211
+ "kinematic_grasp_stability_index": 0.9602531035366528,
212
+ "mean_kinematic_grasp_stability": 0.5547285586085481,
213
+ "per_object_best_kinematic_grasp_stability": {
214
+ "obj_200_cleaner_2": 0.9018268072067874,
215
+ "obj_209_soap_3": 0.8462855464369575
216
+ },
217
+ "grasp_gsi_diagnostics": {
218
+ "mean_anti_slip_score": 0.7435844015695166,
219
+ "mean_duration_score": 0.4409994814779304,
220
+ "mean_hold_score": 0.42710296547504034,
221
+ "mean_kinematic_stability_score": 0.129677032841829,
222
+ "mean_legacy_gsi": 0.4863957525206952,
223
+ "mean_lift_score": 0.4777787184173804,
224
+ "mean_motion_stability_score": 0.39553651528583583
225
+ }
226
+ },
227
+ "tool": {
228
+ "tool_selection_accuracy": null,
229
+ "tool_switch_success_rate": null,
230
+ "tool_switch_total": 0,
231
+ "tool_switch_successful": 0
232
+ },
233
+ "runtime": {
234
+ "avg_steps": 1576.82,
235
+ "error_count": 0
236
+ }
237
+ }
238
+ }
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/09_cleaner_moisturizer_box_loading/None/summary.json ADDED
@@ -0,0 +1,225 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 100900000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 45,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 1926,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.9,
50
+ "success_rate_ci": {
51
+ "mean": 0.9,
52
+ "ci_low": 0.82,
53
+ "ci_high": 0.98
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.95
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 23.754074074074005
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.9,
64
+ "hard_violation_rate": 0.0,
65
+ "drop_rate": 0.0,
66
+ "high_speed_violation_rate": 0.0,
67
+ "safety_violation_step_rate": 0.0
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.9,
74
+ "at_end_success_rate": 0.9,
75
+ "task_family_success_rate": {
76
+ "rigid_pick_place": 0.9
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.94,
81
+ "current_chain_depth_progress_score": 0.95,
82
+ "latched_chain_depth_progress_score": 0.95,
83
+ "avg_current_chain_depth": 1.9,
84
+ "avg_latched_chain_depth": 1.9
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9068783680714388,
88
+ "expert_time_s": 21.383333333333333,
89
+ "expert_time_step": 1283,
90
+ "avg_steps_to_success": 1425.2444444444445,
91
+ "avg_policy_steps_to_success": 475.4222222222222,
92
+ "avg_policy_queries_to_success": 30.155555555555555
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0,
96
+ "violation_counts": {},
97
+ "robot_constraint_violation_rate": 1.0,
98
+ "finger_joint_limit_saturation_rate": 0.96,
99
+ "joint_limit_active_step_rate": 0.10521107285199144,
100
+ "finger_joint_limit_saturation_step_rate": 0.059864978445354225,
101
+ "max_joint_limit_excess_rad": 1.1146379709243774,
102
+ "robot_constraint_diagnostics": {
103
+ "joint_limit_soft_counts": {
104
+ "leap_l_10": 1416,
105
+ "leap_l_2": 368,
106
+ "multi_leap_r_1": 746,
107
+ "multi_leap_r_10": 2921,
108
+ "multi_leap_r_12": 39,
109
+ "multi_leap_r_13": 506,
110
+ "multi_leap_r_2": 1789,
111
+ "multi_leap_r_5": 934,
112
+ "multi_leap_r_6": 579,
113
+ "multi_leap_r_9": 2070
114
+ },
115
+ "joint_limit_hard_counts": {
116
+ "leap_l_10": 56,
117
+ "leap_l_11": 1,
118
+ "leap_l_13": 1,
119
+ "leap_l_2": 330,
120
+ "multi_leap_r_1": 29,
121
+ "multi_leap_r_10": 207,
122
+ "multi_leap_r_12": 3,
123
+ "multi_leap_r_13": 9,
124
+ "multi_leap_r_2": 101,
125
+ "multi_leap_r_5": 27,
126
+ "multi_leap_r_6": 13,
127
+ "multi_leap_r_9": 45
128
+ },
129
+ "joint_limit_hard_episode_events": 608,
130
+ "finger_joint_limit_saturation_counts": {
131
+ "leap_l_10": 900,
132
+ "multi_leap_r_1": 198,
133
+ "multi_leap_r_10": 1762,
134
+ "multi_leap_r_12": 9,
135
+ "multi_leap_r_13": 450,
136
+ "multi_leap_r_2": 1032,
137
+ "multi_leap_r_5": 346,
138
+ "multi_leap_r_6": 302,
139
+ "multi_leap_r_9": 889
140
+ },
141
+ "joint_limit_max_excess_by_joint": {
142
+ "leap_l_10": 1.1146379709243774,
143
+ "leap_l_11": 0.007532387971878052,
144
+ "leap_l_13": 0.006947755813598633,
145
+ "leap_l_15": 0.0003148317337036133,
146
+ "leap_l_2": 0.1518210768699646,
147
+ "leap_l_3": 8.752942085266113e-05,
148
+ "multi_leap_r_1": 0.369361937046051,
149
+ "multi_leap_r_10": 0.6687592267990112,
150
+ "multi_leap_r_11": 0.003950238227844238,
151
+ "multi_leap_r_12": 0.45833802223205566,
152
+ "multi_leap_r_13": 0.7840425968170166,
153
+ "multi_leap_r_2": 0.6488338708877563,
154
+ "multi_leap_r_3": 0.001452118158340454,
155
+ "multi_leap_r_5": 0.39125531911849976,
156
+ "multi_leap_r_6": 0.5104714632034302,
157
+ "multi_leap_r_7": 0.0022551119327545166,
158
+ "multi_leap_r_9": 0.5798805356025696
159
+ },
160
+ "coverage": {
161
+ "hard_limit_source": {
162
+ "available": 50,
163
+ "unavailable": 0,
164
+ "unknown": 0
165
+ },
166
+ "joint_role_manifest": {
167
+ "available": 50,
168
+ "unavailable": 0,
169
+ "unknown": 0
170
+ }
171
+ },
172
+ "hard_limit_sources": {
173
+ "runtime.joint_pos_limits": 50
174
+ },
175
+ "soft_limit_sources": {
176
+ "runtime.soft_joint_pos_limits": 50
177
+ },
178
+ "unknown_runtime_joint_names": [],
179
+ "baseline_checks": {
180
+ "post_home": {
181
+ "checked": 50,
182
+ "hard_violation_episodes": 0
183
+ },
184
+ "post_reset": {
185
+ "checked": 50,
186
+ "hard_violation_episodes": 0
187
+ }
188
+ }
189
+ }
190
+ },
191
+ "robot_motion": {
192
+ "joint_acceleration_rms": 19.797336649475405,
193
+ "joint_effort_rms": 65.3054474958859,
194
+ "joint_jerk_rms": 1977.1634308368941,
195
+ "joint_velocity_rms": 0.6854916393582621
196
+ },
197
+ "grasp": {
198
+ "kinematic_grasp_stability_index": 0.970468080496175,
199
+ "mean_kinematic_grasp_stability": 0.4848565176231215,
200
+ "per_object_best_kinematic_grasp_stability": {
201
+ "obj_200_cleaner_2": 0.8964569143900374,
202
+ "obj_209_soap_3": 0.9184099551051745
203
+ },
204
+ "grasp_gsi_diagnostics": {
205
+ "mean_anti_slip_score": 0.7549050411822381,
206
+ "mean_duration_score": 0.41023123918760973,
207
+ "mean_hold_score": 0.39482209182797706,
208
+ "mean_kinematic_stability_score": 0.10651466258862081,
209
+ "mean_legacy_gsi": 0.4648275389664774,
210
+ "mean_lift_score": 0.4583300713733414,
211
+ "mean_motion_stability_score": 0.3657735919088936
212
+ }
213
+ },
214
+ "tool": {
215
+ "tool_selection_accuracy": null,
216
+ "tool_switch_success_rate": null,
217
+ "tool_switch_total": 0,
218
+ "tool_switch_successful": 0
219
+ },
220
+ "runtime": {
221
+ "avg_steps": 1475.32,
222
+ "error_count": 0
223
+ }
224
+ }
225
+ }
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Equi/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Equi/summary.json ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 101200000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 29,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3447,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.58,
50
+ "success_rate_ci": {
51
+ "mean": 0.58,
52
+ "ci_low": 0.44,
53
+ "ci_high": 0.72
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.78
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 39.42011494252778
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.36,
64
+ "hard_violation_rate": 0.48,
65
+ "drop_rate": 0.1,
66
+ "high_speed_violation_rate": 0.42,
67
+ "safety_violation_step_rate": 0.0003262920455673935
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.6,
74
+ "at_end_success_rate": 0.58,
75
+ "task_family_success_rate": {
76
+ "tool_precision_place": 0.58
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.77,
81
+ "current_chain_depth_progress_score": 0.775,
82
+ "latched_chain_depth_progress_score": 0.78,
83
+ "avg_current_chain_depth": 3.1,
84
+ "avg_latched_chain_depth": 3.12
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9748069140644617,
88
+ "expert_time_s": 38.28333333333333,
89
+ "expert_time_step": 2297,
90
+ "avg_steps_to_success": 2365.206896551724,
91
+ "avg_policy_steps_to_success": 788.7586206896551,
92
+ "avg_policy_queries_to_success": 49.724137931034484
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.00022698577082949113,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_043_phillips_screwdriver_1": 3,
98
+ "drop_to_table_obj_048_hammer_3": 2,
99
+ "high_speed_obj_043_phillips_screwdriver_1": 13,
100
+ "high_speed_obj_044_flat_screwdriver_2": 2,
101
+ "high_speed_obj_048_hammer_3": 12
102
+ },
103
+ "robot_constraint_violation_rate": 1.0,
104
+ "finger_joint_limit_saturation_rate": 1.0,
105
+ "joint_limit_active_step_rate": 0.1832910099448141,
106
+ "finger_joint_limit_saturation_step_rate": 0.13702847252762843,
107
+ "max_joint_limit_excess_rad": 2.0894558429718018,
108
+ "robot_constraint_diagnostics": {
109
+ "joint_limit_soft_counts": {
110
+ "left_index_1_joint": 3312,
111
+ "left_little_1_joint": 5625,
112
+ "left_little_2_joint": 30,
113
+ "left_middle_1_joint": 7324,
114
+ "left_middle_2_joint": 17,
115
+ "left_ring_1_joint": 7457,
116
+ "left_ring_2_joint": 11,
117
+ "left_thumb_2_joint": 3103,
118
+ "left_thumb_3_joint": 2777,
119
+ "left_thumb_4_joint": 3665,
120
+ "right_index_1_joint": 7139,
121
+ "right_little_1_joint": 2970,
122
+ "right_middle_1_joint": 7144,
123
+ "right_middle_2_joint": 23,
124
+ "right_ring_1_joint": 5125,
125
+ "right_ring_2_joint": 39,
126
+ "right_thumb_2_joint": 442,
127
+ "right_thumb_3_joint": 448,
128
+ "right_thumb_4_joint": 1172
129
+ },
130
+ "joint_limit_hard_counts": {
131
+ "left_index_1_joint": 123,
132
+ "left_index_2_joint": 1,
133
+ "left_little_1_joint": 125,
134
+ "left_little_2_joint": 56,
135
+ "left_middle_1_joint": 137,
136
+ "left_middle_2_joint": 37,
137
+ "left_ring_1_joint": 136,
138
+ "left_ring_2_joint": 48,
139
+ "left_thumb_2_joint": 118,
140
+ "left_thumb_3_joint": 51,
141
+ "left_thumb_4_joint": 127,
142
+ "right_index_1_joint": 354,
143
+ "right_little_1_joint": 59,
144
+ "right_middle_1_joint": 119,
145
+ "right_middle_2_joint": 22,
146
+ "right_ring_1_joint": 114,
147
+ "right_ring_2_joint": 38,
148
+ "right_thumb_2_joint": 14,
149
+ "right_thumb_3_joint": 39,
150
+ "right_thumb_4_joint": 56
151
+ },
152
+ "joint_limit_hard_episode_events": 607,
153
+ "finger_joint_limit_saturation_counts": {
154
+ "left_index_1_joint": 1601,
155
+ "left_little_1_joint": 3825,
156
+ "left_middle_1_joint": 5115,
157
+ "left_ring_1_joint": 5453,
158
+ "left_thumb_2_joint": 1431,
159
+ "left_thumb_3_joint": 2151,
160
+ "left_thumb_4_joint": 2562,
161
+ "right_index_1_joint": 4416,
162
+ "right_little_1_joint": 2229,
163
+ "right_middle_1_joint": 4870,
164
+ "right_ring_1_joint": 3216,
165
+ "right_thumb_2_joint": 48,
166
+ "right_thumb_3_joint": 231,
167
+ "right_thumb_4_joint": 595
168
+ },
169
+ "joint_limit_max_excess_by_joint": {
170
+ "left_index_1_joint": 0.711759090423584,
171
+ "left_index_2_joint": 0.005880043376237154,
172
+ "left_little_1_joint": 1.1162053346633911,
173
+ "left_little_2_joint": 0.03149653226137161,
174
+ "left_middle_1_joint": 1.13613760471344,
175
+ "left_middle_2_joint": 0.02122342586517334,
176
+ "left_ring_1_joint": 1.0663917064666748,
177
+ "left_ring_2_joint": 0.023791518062353134,
178
+ "left_thumb_1_joint": 2.262142334075179e-05,
179
+ "left_thumb_2_joint": 0.4295888841152191,
180
+ "left_thumb_3_joint": 1.2575397491455078,
181
+ "left_thumb_4_joint": 1.4528605937957764,
182
+ "right_index_1_joint": 1.0422031879425049,
183
+ "right_index_2_joint": 0.0012161345221102238,
184
+ "right_little_1_joint": 1.5323597192764282,
185
+ "right_little_2_joint": 0.0007862807251513004,
186
+ "right_middle_1_joint": 1.1124267578125,
187
+ "right_middle_2_joint": 0.01797790266573429,
188
+ "right_ring_1_joint": 2.0894558429718018,
189
+ "right_ring_2_joint": 0.03715495765209198,
190
+ "right_thumb_2_joint": 0.1909545660018921,
191
+ "right_thumb_3_joint": 0.8379172682762146,
192
+ "right_thumb_4_joint": 0.7458809614181519
193
+ },
194
+ "coverage": {
195
+ "hard_limit_source": {
196
+ "available": 50,
197
+ "unavailable": 0,
198
+ "unknown": 0
199
+ },
200
+ "joint_role_manifest": {
201
+ "available": 50,
202
+ "unavailable": 0,
203
+ "unknown": 0
204
+ }
205
+ },
206
+ "hard_limit_sources": {
207
+ "runtime.joint_pos_limits": 50
208
+ },
209
+ "soft_limit_sources": {
210
+ "runtime.soft_joint_pos_limits": 50
211
+ },
212
+ "unknown_runtime_joint_names": [],
213
+ "baseline_checks": {
214
+ "post_home": {
215
+ "checked": 50,
216
+ "hard_violation_episodes": 0
217
+ },
218
+ "post_reset": {
219
+ "checked": 50,
220
+ "hard_violation_episodes": 0
221
+ }
222
+ }
223
+ }
224
+ },
225
+ "robot_motion": {
226
+ "joint_acceleration_rms": 60.72977977280714,
227
+ "joint_effort_rms": 50.44087544466842,
228
+ "joint_jerk_rms": 6242.022126320296,
229
+ "joint_velocity_rms": 1.2348019407548574
230
+ },
231
+ "grasp": {
232
+ "kinematic_grasp_stability_index": 0.9760688292429316,
233
+ "mean_kinematic_grasp_stability": 0.45711600712101574,
234
+ "per_object_best_kinematic_grasp_stability": {
235
+ "obj_043_phillips_screwdriver_1": 0.7748434880067786,
236
+ "obj_044_flat_screwdriver_2": 0.893077211204318,
237
+ "obj_048_hammer_3": 0.6322258711314805
238
+ },
239
+ "grasp_gsi_diagnostics": {
240
+ "mean_anti_slip_score": 0.7190484308073894,
241
+ "mean_duration_score": 0.47856261444306986,
242
+ "mean_hold_score": 0.4318505917562669,
243
+ "mean_kinematic_stability_score": 0.16567862359695776,
244
+ "mean_legacy_gsi": 0.49913792700145787,
245
+ "mean_lift_score": 0.5560769556413924,
246
+ "mean_motion_stability_score": 0.41395149634121925
247
+ }
248
+ },
249
+ "tool": {
250
+ "tool_selection_accuracy": 0.0,
251
+ "tool_switch_success_rate": 0.0,
252
+ "tool_switch_total": 153,
253
+ "tool_switch_successful": 0
254
+ },
255
+ "runtime": {
256
+ "avg_steps": 2819.56,
257
+ "error_count": 0
258
+ }
259
+ }
260
+ }
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Full/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Full/summary.json ADDED
@@ -0,0 +1,259 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 201200000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 37,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3447,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.74,
50
+ "success_rate_ci": {
51
+ "mean": 0.74,
52
+ "ci_low": 0.62,
53
+ "ci_high": 0.86
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.835
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 38.58468468468378
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.48,
64
+ "hard_violation_rate": 0.46,
65
+ "drop_rate": 0.1,
66
+ "high_speed_violation_rate": 0.36,
67
+ "safety_violation_step_rate": 0.00031425089484858474
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.74,
74
+ "at_end_success_rate": 0.74,
75
+ "task_family_success_rate": {
76
+ "tool_precision_place": 0.74
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.83,
81
+ "current_chain_depth_progress_score": 0.83,
82
+ "latched_chain_depth_progress_score": 0.835,
83
+ "avg_current_chain_depth": 3.32,
84
+ "avg_latched_chain_depth": 3.34
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9960691517823889,
88
+ "expert_time_s": 38.28333333333333,
89
+ "expert_time_step": 2297,
90
+ "avg_steps_to_success": 2315.0810810810813,
91
+ "avg_policy_steps_to_success": 772.0540540540541,
92
+ "avg_policy_queries_to_success": 48.567567567567565
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.00020694571124175092,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_048_hammer_3": 5,
98
+ "high_speed_obj_043_phillips_screwdriver_1": 14,
99
+ "high_speed_obj_044_flat_screwdriver_2": 1,
100
+ "high_speed_obj_048_hammer_3": 7
101
+ },
102
+ "robot_constraint_violation_rate": 1.0,
103
+ "finger_joint_limit_saturation_rate": 1.0,
104
+ "joint_limit_active_step_rate": 0.18637377461312649,
105
+ "finger_joint_limit_saturation_step_rate": 0.14447876507062982,
106
+ "max_joint_limit_excess_rad": 2.3060953617095947,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "left_index_1_joint": 3717,
110
+ "left_little_1_joint": 6481,
111
+ "left_little_2_joint": 44,
112
+ "left_middle_1_joint": 8151,
113
+ "left_middle_2_joint": 5,
114
+ "left_ring_1_joint": 7891,
115
+ "left_ring_2_joint": 13,
116
+ "left_thumb_2_joint": 4419,
117
+ "left_thumb_3_joint": 2834,
118
+ "left_thumb_4_joint": 4976,
119
+ "right_index_1_joint": 8379,
120
+ "right_little_1_joint": 2959,
121
+ "right_middle_1_joint": 4613,
122
+ "right_middle_2_joint": 9,
123
+ "right_ring_1_joint": 3349,
124
+ "right_ring_2_joint": 66,
125
+ "right_thumb_2_joint": 262,
126
+ "right_thumb_3_joint": 299,
127
+ "right_thumb_4_joint": 858
128
+ },
129
+ "joint_limit_hard_counts": {
130
+ "left_index_1_joint": 122,
131
+ "left_little_1_joint": 145,
132
+ "left_little_2_joint": 70,
133
+ "left_middle_1_joint": 145,
134
+ "left_middle_2_joint": 36,
135
+ "left_ring_1_joint": 147,
136
+ "left_ring_2_joint": 36,
137
+ "left_thumb_2_joint": 135,
138
+ "left_thumb_3_joint": 53,
139
+ "left_thumb_4_joint": 144,
140
+ "right_index_1_joint": 348,
141
+ "right_index_2_joint": 1,
142
+ "right_little_1_joint": 62,
143
+ "right_little_2_joint": 2,
144
+ "right_middle_1_joint": 83,
145
+ "right_middle_2_joint": 15,
146
+ "right_ring_1_joint": 80,
147
+ "right_ring_2_joint": 35,
148
+ "right_thumb_2_joint": 16,
149
+ "right_thumb_3_joint": 37,
150
+ "right_thumb_4_joint": 61
151
+ },
152
+ "joint_limit_hard_episode_events": 554,
153
+ "finger_joint_limit_saturation_counts": {
154
+ "left_index_1_joint": 1762,
155
+ "left_little_1_joint": 4464,
156
+ "left_middle_1_joint": 6102,
157
+ "left_ring_1_joint": 6033,
158
+ "left_thumb_2_joint": 2235,
159
+ "left_thumb_3_joint": 2150,
160
+ "left_thumb_4_joint": 3671,
161
+ "right_index_1_joint": 5529,
162
+ "right_little_1_joint": 2031,
163
+ "right_middle_1_joint": 3306,
164
+ "right_ring_1_joint": 1977,
165
+ "right_thumb_2_joint": 25,
166
+ "right_thumb_3_joint": 149,
167
+ "right_thumb_4_joint": 429
168
+ },
169
+ "joint_limit_max_excess_by_joint": {
170
+ "left_index_1_joint": 0.9117370843887329,
171
+ "left_index_2_joint": 1.2340217381279217e-06,
172
+ "left_little_1_joint": 0.8457517623901367,
173
+ "left_little_2_joint": 0.05706056207418442,
174
+ "left_middle_1_joint": 0.9888004064559937,
175
+ "left_middle_2_joint": 0.0166944433003664,
176
+ "left_ring_1_joint": 0.8540922999382019,
177
+ "left_ring_2_joint": 0.022290682420134544,
178
+ "left_thumb_2_joint": 0.45075151324272156,
179
+ "left_thumb_3_joint": 1.12669038772583,
180
+ "left_thumb_4_joint": 1.4544601440429688,
181
+ "right_index_1_joint": 0.5311816930770874,
182
+ "right_index_2_joint": 0.008225003257393837,
183
+ "right_little_1_joint": 1.5824559926986694,
184
+ "right_little_2_joint": 0.007416172418743372,
185
+ "right_middle_1_joint": 1.0047886371612549,
186
+ "right_middle_2_joint": 0.028083227574825287,
187
+ "right_ring_1_joint": 2.3060953617095947,
188
+ "right_ring_2_joint": 0.05318034067749977,
189
+ "right_thumb_2_joint": 0.24802277982234955,
190
+ "right_thumb_3_joint": 0.6289350390434265,
191
+ "right_thumb_4_joint": 0.588646650314331
192
+ },
193
+ "coverage": {
194
+ "hard_limit_source": {
195
+ "available": 50,
196
+ "unavailable": 0,
197
+ "unknown": 0
198
+ },
199
+ "joint_role_manifest": {
200
+ "available": 50,
201
+ "unavailable": 0,
202
+ "unknown": 0
203
+ }
204
+ },
205
+ "hard_limit_sources": {
206
+ "runtime.joint_pos_limits": 50
207
+ },
208
+ "soft_limit_sources": {
209
+ "runtime.soft_joint_pos_limits": 50
210
+ },
211
+ "unknown_runtime_joint_names": [],
212
+ "baseline_checks": {
213
+ "post_home": {
214
+ "checked": 50,
215
+ "hard_violation_episodes": 0
216
+ },
217
+ "post_reset": {
218
+ "checked": 50,
219
+ "hard_violation_episodes": 0
220
+ }
221
+ }
222
+ }
223
+ },
224
+ "robot_motion": {
225
+ "joint_acceleration_rms": 62.40365783151193,
226
+ "joint_effort_rms": 51.701171834225846,
227
+ "joint_jerk_rms": 6434.713121315065,
228
+ "joint_velocity_rms": 1.2711213855440726
229
+ },
230
+ "grasp": {
231
+ "kinematic_grasp_stability_index": 0.9758896993734455,
232
+ "mean_kinematic_grasp_stability": 0.504071943057219,
233
+ "per_object_best_kinematic_grasp_stability": {
234
+ "obj_043_phillips_screwdriver_1": 0.8690254583212103,
235
+ "obj_044_flat_screwdriver_2": 0.7589282566109706,
236
+ "obj_048_hammer_3": 0.6760367886939807
237
+ },
238
+ "grasp_gsi_diagnostics": {
239
+ "mean_anti_slip_score": 0.7724433985880149,
240
+ "mean_duration_score": 0.4427087894428896,
241
+ "mean_hold_score": 0.39069043678683923,
242
+ "mean_kinematic_stability_score": 0.1784396631680777,
243
+ "mean_legacy_gsi": 0.48022413397585123,
244
+ "mean_lift_score": 0.5216040789795737,
245
+ "mean_motion_stability_score": 0.3762954569376321
246
+ }
247
+ },
248
+ "tool": {
249
+ "tool_selection_accuracy": 0.0,
250
+ "tool_switch_success_rate": 0.0,
251
+ "tool_switch_total": 155,
252
+ "tool_switch_successful": 0
253
+ },
254
+ "runtime": {
255
+ "avg_steps": 2609.38,
256
+ "error_count": 0
257
+ }
258
+ }
259
+ }
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Inv/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/Inv/summary.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metric_schema_version": "core-v1.8-policy-step-query-split",
3
+ "metadata": {
4
+ "policy_name": "ME-Dex 1.5",
5
+ "base_seed": 101200000,
6
+ "seed_policy": "episode_seed = base_seed + episode_index - 1",
7
+ "randomization_recorded": true,
8
+ "n_episodes": 50,
9
+ "successes": 34,
10
+ "error_count": 0,
11
+ "evaluation_protocol": "reach_and_stop",
12
+ "early_stop": true,
13
+ "success_definition": "terminal.raw_condition holds for dwell_time_s",
14
+ "measurement_window": "episode_start_to_stable_success_or_failure_budget",
15
+ "max_episode_steps": 3447,
16
+ "dwell_time_s": 0.5,
17
+ "field_semantics": {
18
+ "success_rate": "primary protocol success rate: stable_success for reach_and_stop; terminal at_end_success_budget for fixed_horizon",
19
+ "avg_time_to_success_s": "main efficiency metric: mean time_to_stable_success_s over reached-success episodes",
20
+ "task_efficiency": "benchmark-specific expert-normalized speed ratio: expert_time_s / time_to_stable_success_s over reached-success episodes",
21
+ "avg_steps_to_success": "diagnostic mean steps_to_stable_success over reached-success episodes, or None if no episode succeeds",
22
+ "avg_policy_steps_to_success": "diagnostic mean policy-rate control/action steps through stable success over reached-success episodes",
23
+ "avg_policy_queries_to_success": "diagnostic mean actual model inference queries through stable success over reached-success episodes; None when the first-success query count was not observed",
24
+ "safe_success_rate": "episode success rate with no task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
25
+ "hard_violation_rate": "episode rate with any task-level hard safety violation (drop or tracked-object high speed; joint limits are diagnostic by default)",
26
+ "drop_rate": "episode rate with drop_off_table or drop_to_table_after_lift violation",
27
+ "high_speed_violation_rate": "episode rate with any tracked object exceeding the configured speed threshold",
28
+ "excessive_impact_rate": "deprecated contact-force impact metric; no longer populated by core safety",
29
+ "robot_constraint_violation_rate": "diagnostic episode rate with any asset hard-joint-limit observation; excluded from safe_success_rate unless include_in_core_safety is explicitly enabled",
30
+ "finger_joint_limit_saturation_rate": "episode rate with dexterous finger joints saturated beyond finger joint-limit thresholds; diagnostic by default and not included in safe_success_rate",
31
+ "joint_limit_active_step_rate": "diagnostic mean fraction of steps with any joint exceeding the soft diagnostic margin",
32
+ "finger_joint_limit_saturation_step_rate": "diagnostic mean fraction of steps with finger joint-limit saturation",
33
+ "max_joint_limit_excess_rad": "diagnostic maximum operating/soft-limit excess observed in radians; hard-limit source and role validation are in robot_constraint_diagnostics",
34
+ "safety_violation_rate_time": "compatibility diagnostic; with fixed physics_dt this is equivalent to safety_violation_step_rate and is omitted from grouped summary output",
35
+ "safety_violation_step_rate": "core safety metric: fraction of executed steps with task-level safety violations; joint-limit observations are excluded by default",
36
+ "safety_violation_rate_event": "deprecated compatibility alias of safety_violation_events_per_step; omitted from grouped summary output",
37
+ "safety_violation_events_per_step": "diagnostic event density; may exceed 1 when multiple violations happen in one step",
38
+ "episode_violation_rate": "compatibility alias of hard_violation_rate; omitted from grouped core output",
39
+ "latched_stage_completion_rate": "main progress metric; ever-reached stage completion rate at episode end",
40
+ "current_stage_completion_rate": "diagnostic terminal-state stage completion rate at episode end",
41
+ "current_normalized_progress_score": "chain-depth diagnostic; normalized progress score at episode end",
42
+ "current_chain_depth_progress_score": "alias of current_normalized_progress_score for ordered dependency-depth diagnostics",
43
+ "latched_chain_depth_progress_score": "alias of latched_normalized_progress_score for ever-reached dependency-depth diagnostics",
44
+ "robustness": "success rates grouped by benchmark protocol category: cov, inv, and inv+cov"
45
+ }
46
+ },
47
+ "core_metrics": {
48
+ "completion": {
49
+ "success_rate": 0.68,
50
+ "success_rate_ci": {
51
+ "mean": 0.68,
52
+ "ci_low": 0.54,
53
+ "ci_high": 0.8
54
+ }
55
+ },
56
+ "progress": {
57
+ "latched_stage_completion_rate": 0.815
58
+ },
59
+ "efficiency": {
60
+ "avg_time_to_success_s": 38.67696078431281
61
+ },
62
+ "safety": {
63
+ "safe_success_rate": 0.46,
64
+ "hard_violation_rate": 0.36,
65
+ "drop_rate": 0.06,
66
+ "high_speed_violation_rate": 0.32,
67
+ "safety_violation_step_rate": 0.0003133089151305827
68
+ },
69
+ "robustness": {}
70
+ },
71
+ "diagnostic_metrics": {
72
+ "completion": {
73
+ "ever_instant_success_rate": 0.72,
74
+ "at_end_success_rate": 0.68,
75
+ "task_family_success_rate": {
76
+ "tool_precision_place": 0.68
77
+ }
78
+ },
79
+ "progress": {
80
+ "current_stage_completion_rate": 0.8,
81
+ "current_chain_depth_progress_score": 0.805,
82
+ "latched_chain_depth_progress_score": 0.815,
83
+ "avg_current_chain_depth": 3.22,
84
+ "avg_latched_chain_depth": 3.26
85
+ },
86
+ "efficiency": {
87
+ "task_efficiency": 0.9965830653363752,
88
+ "expert_time_s": 38.28333333333333,
89
+ "expert_time_step": 2297,
90
+ "avg_steps_to_success": 2320.6176470588234,
91
+ "avg_policy_steps_to_success": 773.8529411764706,
92
+ "avg_policy_queries_to_success": 48.705882352941174
93
+ },
94
+ "safety": {
95
+ "safety_violation_events_per_step": 0.0001715739297143667,
96
+ "violation_counts": {
97
+ "drop_to_table_obj_048_hammer_3": 3,
98
+ "high_speed_obj_043_phillips_screwdriver_1": 10,
99
+ "high_speed_obj_044_flat_screwdriver_2": 2,
100
+ "high_speed_obj_048_hammer_3": 8
101
+ },
102
+ "robot_constraint_violation_rate": 1.0,
103
+ "finger_joint_limit_saturation_rate": 1.0,
104
+ "joint_limit_active_step_rate": 0.19080512931452484,
105
+ "finger_joint_limit_saturation_step_rate": 0.15303648556914057,
106
+ "max_joint_limit_excess_rad": 1.9719761610031128,
107
+ "robot_constraint_diagnostics": {
108
+ "joint_limit_soft_counts": {
109
+ "left_index_1_joint": 3369,
110
+ "left_little_1_joint": 6330,
111
+ "left_little_2_joint": 40,
112
+ "left_middle_1_joint": 7560,
113
+ "left_middle_2_joint": 9,
114
+ "left_ring_1_joint": 7393,
115
+ "left_ring_2_joint": 26,
116
+ "left_thumb_2_joint": 3391,
117
+ "left_thumb_3_joint": 3402,
118
+ "left_thumb_4_joint": 4599,
119
+ "right_index_1_joint": 9415,
120
+ "right_index_2_joint": 38,
121
+ "right_little_1_joint": 3244,
122
+ "right_middle_1_joint": 6045,
123
+ "right_middle_2_joint": 83,
124
+ "right_ring_1_joint": 3922,
125
+ "right_ring_2_joint": 253,
126
+ "right_thumb_1_joint": 26,
127
+ "right_thumb_2_joint": 479,
128
+ "right_thumb_3_joint": 575,
129
+ "right_thumb_4_joint": 1176
130
+ },
131
+ "joint_limit_hard_counts": {
132
+ "left_index_1_joint": 132,
133
+ "left_little_1_joint": 135,
134
+ "left_little_2_joint": 75,
135
+ "left_middle_1_joint": 133,
136
+ "left_middle_2_joint": 46,
137
+ "left_ring_1_joint": 132,
138
+ "left_ring_2_joint": 66,
139
+ "left_thumb_2_joint": 149,
140
+ "left_thumb_3_joint": 61,
141
+ "left_thumb_4_joint": 169,
142
+ "right_index_1_joint": 307,
143
+ "right_index_2_joint": 5,
144
+ "right_little_1_joint": 61,
145
+ "right_middle_1_joint": 88,
146
+ "right_middle_2_joint": 39,
147
+ "right_ring_1_joint": 92,
148
+ "right_ring_2_joint": 46,
149
+ "right_thumb_1_joint": 3,
150
+ "right_thumb_2_joint": 21,
151
+ "right_thumb_3_joint": 19,
152
+ "right_thumb_4_joint": 69
153
+ },
154
+ "joint_limit_hard_episode_events": 485,
155
+ "finger_joint_limit_saturation_counts": {
156
+ "left_index_1_joint": 1363,
157
+ "left_little_1_joint": 4379,
158
+ "left_middle_1_joint": 5609,
159
+ "left_ring_1_joint": 5571,
160
+ "left_thumb_2_joint": 1480,
161
+ "left_thumb_3_joint": 2498,
162
+ "left_thumb_4_joint": 3350,
163
+ "right_index_1_joint": 6434,
164
+ "right_little_1_joint": 2178,
165
+ "right_middle_1_joint": 4352,
166
+ "right_ring_1_joint": 2428,
167
+ "right_ring_2_joint": 186,
168
+ "right_thumb_2_joint": 60,
169
+ "right_thumb_3_joint": 360,
170
+ "right_thumb_4_joint": 569
171
+ },
172
+ "joint_limit_max_excess_by_joint": {
173
+ "left_index_1_joint": 0.6154660582542419,
174
+ "left_index_2_joint": 6.713170250804978e-07,
175
+ "left_little_1_joint": 0.9938316345214844,
176
+ "left_little_2_joint": 0.06040250137448311,
177
+ "left_middle_1_joint": 0.842703640460968,
178
+ "left_middle_2_joint": 0.016895221546292305,
179
+ "left_ring_1_joint": 0.8439968824386597,
180
+ "left_ring_2_joint": 0.02375255897641182,
181
+ "left_thumb_2_joint": 0.4557880163192749,
182
+ "left_thumb_3_joint": 1.262162208557129,
183
+ "left_thumb_4_joint": 1.5495535135269165,
184
+ "right_index_1_joint": 1.0428470373153687,
185
+ "right_index_2_joint": 0.037547193467617035,
186
+ "right_little_1_joint": 1.0608679056167603,
187
+ "right_little_2_joint": 9.732447506394237e-05,
188
+ "right_middle_1_joint": 1.9719761610031128,
189
+ "right_middle_2_joint": 0.034843917936086655,
190
+ "right_ring_1_joint": 1.9236699342727661,
191
+ "right_ring_2_joint": 0.8722777962684631,
192
+ "right_thumb_1_joint": 0.1128765195608139,
193
+ "right_thumb_2_joint": 0.2366730272769928,
194
+ "right_thumb_3_joint": 0.492065966129303,
195
+ "right_thumb_4_joint": 0.6257126331329346
196
+ },
197
+ "coverage": {
198
+ "hard_limit_source": {
199
+ "available": 50,
200
+ "unavailable": 0,
201
+ "unknown": 0
202
+ },
203
+ "joint_role_manifest": {
204
+ "available": 50,
205
+ "unavailable": 0,
206
+ "unknown": 0
207
+ }
208
+ },
209
+ "hard_limit_sources": {
210
+ "runtime.joint_pos_limits": 50
211
+ },
212
+ "soft_limit_sources": {
213
+ "runtime.soft_joint_pos_limits": 50
214
+ },
215
+ "unknown_runtime_joint_names": [],
216
+ "baseline_checks": {
217
+ "post_home": {
218
+ "checked": 50,
219
+ "hard_violation_episodes": 0
220
+ },
221
+ "post_reset": {
222
+ "checked": 50,
223
+ "hard_violation_episodes": 0
224
+ }
225
+ }
226
+ }
227
+ },
228
+ "robot_motion": {
229
+ "joint_acceleration_rms": 62.283608738427574,
230
+ "joint_effort_rms": 51.399689905348936,
231
+ "joint_jerk_rms": 6401.119316986136,
232
+ "joint_velocity_rms": 1.272511328478383
233
+ },
234
+ "grasp": {
235
+ "kinematic_grasp_stability_index": 0.961276928334697,
236
+ "mean_kinematic_grasp_stability": 0.5009503444573447,
237
+ "per_object_best_kinematic_grasp_stability": {
238
+ "obj_043_phillips_screwdriver_1": 0.8786878235202732,
239
+ "obj_044_flat_screwdriver_2": 0.8466573731336625,
240
+ "obj_048_hammer_3": 0.6403724596178583
241
+ },
242
+ "grasp_gsi_diagnostics": {
243
+ "mean_anti_slip_score": 0.7819799319805003,
244
+ "mean_duration_score": 0.4519653533842103,
245
+ "mean_hold_score": 0.3986048238210461,
246
+ "mean_kinematic_stability_score": 0.19141256407579818,
247
+ "mean_legacy_gsi": 0.48776226415169793,
248
+ "mean_lift_score": 0.5287704236207421,
249
+ "mean_motion_stability_score": 0.3836824564974745
250
+ }
251
+ },
252
+ "tool": {
253
+ "tool_selection_accuracy": 0.0,
254
+ "tool_switch_success_rate": 0.006802721088435374,
255
+ "tool_switch_total": 158,
256
+ "tool_switch_successful": 1
257
+ },
258
+ "runtime": {
259
+ "avg_steps": 2681.06,
260
+ "error_count": 0
261
+ }
262
+ }
263
+ }
me-dex-1-5/per_task_eval/12_screwdriver_box_and_hammer/None/per_episode.jsonl ADDED
The diff for this file is too large to render. See raw diff