yzzhao commited on
Commit
5d7e5dd
·
verified ·
1 Parent(s): 518ecbf

Complete nine RoboTwin behavior episodes with native policy recovery evidence

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +1 -0
  2. data.json +14 -1
  3. experiments/robotwin-native/kinex_policy-3/artifacts.json +109 -0
  4. experiments/robotwin-native/kinex_policy-3/combined.mp4 +3 -0
  5. experiments/robotwin-native/kinex_policy-3/detail.json +0 -0
  6. experiments/robotwin-native/kinex_policy-3/effective-model.json +16 -0
  7. experiments/robotwin-native/kinex_policy-3/episode-report.json +540 -0
  8. experiments/robotwin-native/kinex_policy-3/episode.json +15 -0
  9. experiments/robotwin-native/kinex_policy-3/evidence.tar.gz +3 -0
  10. experiments/robotwin-native/kinex_policy-3/instruction.json +14 -0
  11. experiments/robotwin-native/kinex_policy-3/media.json +101 -0
  12. experiments/robotwin-native/kinex_policy-3/observations/0b176341-1.png +0 -0
  13. experiments/robotwin-native/kinex_policy-3/observations/13ff48a6-1.png +0 -0
  14. experiments/robotwin-native/kinex_policy-3/observations/173a92e8-1.png +0 -0
  15. experiments/robotwin-native/kinex_policy-3/observations/1d570650-1.png +0 -0
  16. experiments/robotwin-native/kinex_policy-3/observations/265cda94-1.png +0 -0
  17. experiments/robotwin-native/kinex_policy-3/observations/27bf832e-1.png +0 -0
  18. experiments/robotwin-native/kinex_policy-3/observations/2cb2e601-1.png +0 -0
  19. experiments/robotwin-native/kinex_policy-3/observations/2ced7a9d-1.png +0 -0
  20. experiments/robotwin-native/kinex_policy-3/observations/36fe7431-1.png +0 -0
  21. experiments/robotwin-native/kinex_policy-3/observations/36fe7431-2.png +0 -0
  22. experiments/robotwin-native/kinex_policy-3/observations/40c24470-1.png +0 -0
  23. experiments/robotwin-native/kinex_policy-3/observations/41da9eb5-1.png +0 -0
  24. experiments/robotwin-native/kinex_policy-3/observations/44263e03-1.png +0 -0
  25. experiments/robotwin-native/kinex_policy-3/observations/6d700e8b-1.png +0 -0
  26. experiments/robotwin-native/kinex_policy-3/observations/6f0ea5a7-1.png +0 -0
  27. experiments/robotwin-native/kinex_policy-3/observations/6f40ac77-1.png +0 -0
  28. experiments/robotwin-native/kinex_policy-3/observations/7ad93479-1.png +0 -0
  29. experiments/robotwin-native/kinex_policy-3/observations/7d1d9cc1-1.png +0 -0
  30. experiments/robotwin-native/kinex_policy-3/observations/7d40da85-1.png +0 -0
  31. experiments/robotwin-native/kinex_policy-3/observations/806acf8c-1.png +0 -0
  32. experiments/robotwin-native/kinex_policy-3/observations/81da75c1-1.png +0 -0
  33. experiments/robotwin-native/kinex_policy-3/observations/82bb4d8a-1.png +0 -0
  34. experiments/robotwin-native/kinex_policy-3/observations/92a89f2b-1.png +0 -0
  35. experiments/robotwin-native/kinex_policy-3/observations/935284a6-1.png +0 -0
  36. experiments/robotwin-native/kinex_policy-3/observations/9ba4e93f-1.png +0 -0
  37. experiments/robotwin-native/kinex_policy-3/observations/9e391af7-1.png +0 -0
  38. experiments/robotwin-native/kinex_policy-3/observations/a119dcde-1.png +0 -0
  39. experiments/robotwin-native/kinex_policy-3/observations/a39bc7a3-1.png +0 -0
  40. experiments/robotwin-native/kinex_policy-3/observations/a85fdbef-1.png +0 -0
  41. experiments/robotwin-native/kinex_policy-3/observations/a8978258-1.png +0 -0
  42. experiments/robotwin-native/kinex_policy-3/observations/b03aa410-1.png +0 -0
  43. experiments/robotwin-native/kinex_policy-3/observations/b0b9d8e3-1.png +0 -0
  44. experiments/robotwin-native/kinex_policy-3/observations/b2b427a8-1.png +0 -0
  45. experiments/robotwin-native/kinex_policy-3/observations/ba87585e-1.png +0 -0
  46. experiments/robotwin-native/kinex_policy-3/observations/c5bb466a-1.png +0 -0
  47. experiments/robotwin-native/kinex_policy-3/observations/c5bb466a-2.png +0 -0
  48. experiments/robotwin-native/kinex_policy-3/observations/d2b3d472-1.png +0 -0
  49. experiments/robotwin-native/kinex_policy-3/observations/d4a17c8a-1.png +0 -0
  50. experiments/robotwin-native/kinex_policy-3/observations/e8c5e432-1.png +0 -0
.gitattributes CHANGED
@@ -62,3 +62,4 @@ experiments/robotwin-native/kinex-2/combined.mp4 filter=lfs diff=lfs merge=lfs -
62
  experiments/robotwin-native/standalone-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
63
  experiments/robotwin-native/kinex_policy-2/combined.mp4 filter=lfs diff=lfs merge=lfs -text
64
  experiments/robotwin-native/kinex-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
 
 
62
  experiments/robotwin-native/standalone-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
63
  experiments/robotwin-native/kinex_policy-2/combined.mp4 filter=lfs diff=lfs merge=lfs -text
64
  experiments/robotwin-native/kinex-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
65
+ experiments/robotwin-native/kinex_policy-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
data.json CHANGED
@@ -196,7 +196,20 @@
196
  "id": "kinex_policy-3",
197
  "task": "3",
198
  "condition": "kinex_policy",
199
- "status": "running"
 
 
 
 
 
 
 
 
 
 
 
 
 
200
  }
201
  ]
202
  }
 
196
  "id": "kinex_policy-3",
197
  "task": "3",
198
  "condition": "kinex_policy",
199
+ "status": "complete",
200
+ "success": true,
201
+ "valid": true,
202
+ "control_steps": 1277,
203
+ "tool_calls": 69,
204
+ "vla_calls": 72,
205
+ "input_tokens": 2083665,
206
+ "output_tokens": 5246,
207
+ "cached_input_tokens": 2026752,
208
+ "api_cost_usd": 2.858182,
209
+ "path": "experiments/robotwin-native/kinex_policy-3",
210
+ "video": "experiments/robotwin-native/kinex_policy-3/combined.mp4",
211
+ "poster": "experiments/robotwin-native/kinex_policy-3/poster.jpg",
212
+ "wall_time_s": 737.794664
213
  }
214
  ]
215
  }
experiments/robotwin-native/kinex_policy-3/artifacts.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "evidence": {
3
+ "complete": true,
4
+ "path": "evidence.tar.gz"
5
+ },
6
+ "recording": {
7
+ "complete": true,
8
+ "enabled": true,
9
+ "encoder_queue_drops": 0,
10
+ "lossless": false,
11
+ "manifest": {
12
+ "bag_messages": 0,
13
+ "error": null,
14
+ "fps": 10.0,
15
+ "run_id": "88c15864ae34",
16
+ "schema": "rosbridge/recording/1",
17
+ "spectator_views": [
18
+ "third_person",
19
+ "left_wrist",
20
+ "right_wrist"
21
+ ],
22
+ "state": "complete",
23
+ "task": "robotwin/put-bottles-dustbin-aloha-agilex",
24
+ "topics": [],
25
+ "type": "mp4",
26
+ "views": {
27
+ "left_wrist": {
28
+ "dropped": 0,
29
+ "error": null,
30
+ "received": 512,
31
+ "segments": [
32
+ {
33
+ "file": "left_wrist.mp4",
34
+ "first_stamp_ns": 0,
35
+ "fps": 10.0,
36
+ "frames": 512,
37
+ "height": 720,
38
+ "last_stamp_ns": 51080000000,
39
+ "width": 960
40
+ }
41
+ ]
42
+ },
43
+ "right_wrist": {
44
+ "dropped": 0,
45
+ "error": null,
46
+ "received": 512,
47
+ "segments": [
48
+ {
49
+ "file": "right_wrist.mp4",
50
+ "first_stamp_ns": 0,
51
+ "fps": 10.0,
52
+ "frames": 512,
53
+ "height": 720,
54
+ "last_stamp_ns": 51080000000,
55
+ "width": 960
56
+ }
57
+ ]
58
+ },
59
+ "third_person": {
60
+ "dropped": 0,
61
+ "error": null,
62
+ "received": 512,
63
+ "segments": [
64
+ {
65
+ "file": "third_person.mp4",
66
+ "first_stamp_ns": 0,
67
+ "fps": 10.0,
68
+ "frames": 512,
69
+ "height": 720,
70
+ "last_stamp_ns": 51080000000,
71
+ "width": 960
72
+ }
73
+ ]
74
+ }
75
+ }
76
+ },
77
+ "video_available": true,
78
+ "videos": [
79
+ {
80
+ "bytes": 4677001,
81
+ "channel": "spectator",
82
+ "fps": 10.0,
83
+ "frames": 512,
84
+ "path": "recording/third_person.mp4"
85
+ },
86
+ {
87
+ "bytes": 5003605,
88
+ "channel": "spectator",
89
+ "fps": 10.0,
90
+ "frames": 512,
91
+ "path": "recording/left_wrist.mp4"
92
+ },
93
+ {
94
+ "bytes": 5726307,
95
+ "channel": "spectator",
96
+ "fps": 10.0,
97
+ "frames": 512,
98
+ "path": "recording/right_wrist.mp4"
99
+ }
100
+ ]
101
+ },
102
+ "schema": "rlebench/episode-artifacts/1",
103
+ "terminal_evidence": {
104
+ "available": false,
105
+ "complete": false,
106
+ "error": "FileNotFoundError: [Errno 2] No such file or directory: '/run/rlebench/private/episode_01/evidence/terminal.json'"
107
+ },
108
+ "workspace_complete": true
109
+ }
experiments/robotwin-native/kinex_policy-3/combined.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9369bf616d6758deb91a356b9195cb4bebcd7e6f796e0d9d853fda219c238d0
3
+ size 9650040
experiments/robotwin-native/kinex_policy-3/detail.json ADDED
The diff for this file is too large to render. See raw diff
 
experiments/robotwin-native/kinex_policy-3/effective-model.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "kinex_identity": "74635879f639d2bf8918b7fab95bf1b37e29c187d24b1ea5692f22079d81d24f",
3
+ "kinex_version": "1.1.0",
4
+ "kinex_source": null,
5
+ "model": "openai-codex/gpt-6-astra",
6
+ "effort": "high",
7
+ "service_tier": null,
8
+ "fresh_session": true,
9
+ "max_steps": null,
10
+ "model_recording": false,
11
+ "compaction": "native",
12
+ "transport": "sse",
13
+ "strict_usage": true,
14
+ "max_request_retries": 50,
15
+ "usage_accounting": "reported-responses"
16
+ }
experiments/robotwin-native/kinex_policy-3/episode-report.json ADDED
@@ -0,0 +1,540 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "episode": 1,
3
+ "initialization": {
4
+ "expert_plan_success": false,
5
+ "requested_seed": 0
6
+ },
7
+ "native_reward": 1.0,
8
+ "planned_episodes": 1,
9
+ "reason": "success",
10
+ "simulation_time_s": 51.08,
11
+ "step_limit": 1700,
12
+ "steps": 1277,
13
+ "success": true,
14
+ "valid": true,
15
+ "step": "episode_01",
16
+ "evidence_valid": true,
17
+ "wall_time_sec": 737.794664,
18
+ "usage": {
19
+ "n_input_tokens": 2083665,
20
+ "n_output_tokens": 5246,
21
+ "n_cache_tokens": 2026752,
22
+ "cost_usd": null,
23
+ "n_fresh_input_tokens": 56913,
24
+ "n_cache_read_tokens": 2026752,
25
+ "n_cache_write_tokens": 0
26
+ },
27
+ "usage_details": {
28
+ "accounting": "reported-responses",
29
+ "audit_complete": true,
30
+ "audit_source": "Provider terminal response usage",
31
+ "cache_hit_rate": 0.9726861083715472,
32
+ "cache_read_input_tokens": 2026752,
33
+ "cache_reported_input_tokens": 2083665,
34
+ "cache_write_basis": "provider-reported",
35
+ "cache_write_input_tokens": 0,
36
+ "cache_write_reported_input_tokens": 2083665,
37
+ "cached_input_tokens": 2026752,
38
+ "cost_usd": null,
39
+ "fresh_input_tokens": 56913,
40
+ "harbor_mismatches": {},
41
+ "input_breakdown_complete": true,
42
+ "input_tokens": 2083665,
43
+ "known_cache_write_input_tokens": 0,
44
+ "known_cached_input_tokens": 2026752,
45
+ "known_fresh_input_tokens": 56913,
46
+ "known_input_tokens": 2083665,
47
+ "known_output_tokens": 5246,
48
+ "known_reasoning_output_tokens": 328,
49
+ "output_tokens": 5246,
50
+ "parser": "harbor/0.22.0",
51
+ "reasoning_output_tokens": 328,
52
+ "reasoning_reported_output_tokens": 5246,
53
+ "rejected_attempts": 0,
54
+ "reported_responses": {
55
+ "cache_reported_input_tokens": 60,
56
+ "cache_write_input_tokens": 60,
57
+ "cache_write_reported_input_tokens": 60,
58
+ "cached_input_tokens": 60,
59
+ "fresh_input_tokens": 60,
60
+ "input_tokens": 60,
61
+ "output_tokens": 60,
62
+ "reasoning_output_tokens": 60,
63
+ "reasoning_reported_output_tokens": 60
64
+ },
65
+ "request_attempts": 60,
66
+ "response_count": 60,
67
+ "response_ids_complete": true,
68
+ "schema": "rlebench/token-usage/1",
69
+ "source": "Harbor session metrics",
70
+ "total_tokens": 2088911,
71
+ "uncached_input_tokens": 56913,
72
+ "unresolved_attempts": [],
73
+ "usage_records": [
74
+ {
75
+ "cache_write_input_tokens": 0,
76
+ "cached_input_tokens": 0,
77
+ "input_tokens": 11303,
78
+ "output_tokens": 19,
79
+ "reasoning_output_tokens": 0
80
+ },
81
+ {
82
+ "cache_write_input_tokens": 0,
83
+ "cached_input_tokens": 11136,
84
+ "input_tokens": 11343,
85
+ "output_tokens": 29,
86
+ "reasoning_output_tokens": 0
87
+ },
88
+ {
89
+ "cache_write_input_tokens": 0,
90
+ "cached_input_tokens": 11264,
91
+ "input_tokens": 11476,
92
+ "output_tokens": 30,
93
+ "reasoning_output_tokens": 0
94
+ },
95
+ {
96
+ "cache_write_input_tokens": 0,
97
+ "cached_input_tokens": 11392,
98
+ "input_tokens": 15499,
99
+ "output_tokens": 180,
100
+ "reasoning_output_tokens": 0
101
+ },
102
+ {
103
+ "cache_write_input_tokens": 0,
104
+ "cached_input_tokens": 15360,
105
+ "input_tokens": 17418,
106
+ "output_tokens": 210,
107
+ "reasoning_output_tokens": 0
108
+ },
109
+ {
110
+ "cache_write_input_tokens": 0,
111
+ "cached_input_tokens": 17280,
112
+ "input_tokens": 22354,
113
+ "output_tokens": 77,
114
+ "reasoning_output_tokens": 10
115
+ },
116
+ {
117
+ "cache_write_input_tokens": 0,
118
+ "cached_input_tokens": 22272,
119
+ "input_tokens": 22900,
120
+ "output_tokens": 81,
121
+ "reasoning_output_tokens": 0
122
+ },
123
+ {
124
+ "cache_write_input_tokens": 0,
125
+ "cached_input_tokens": 22784,
126
+ "input_tokens": 23208,
127
+ "output_tokens": 77,
128
+ "reasoning_output_tokens": 0
129
+ },
130
+ {
131
+ "cache_write_input_tokens": 0,
132
+ "cached_input_tokens": 23040,
133
+ "input_tokens": 24559,
134
+ "output_tokens": 107,
135
+ "reasoning_output_tokens": 0
136
+ },
137
+ {
138
+ "cache_write_input_tokens": 0,
139
+ "cached_input_tokens": 24448,
140
+ "input_tokens": 24883,
141
+ "output_tokens": 23,
142
+ "reasoning_output_tokens": 0
143
+ },
144
+ {
145
+ "cache_write_input_tokens": 0,
146
+ "cached_input_tokens": 24704,
147
+ "input_tokens": 25073,
148
+ "output_tokens": 40,
149
+ "reasoning_output_tokens": 0
150
+ },
151
+ {
152
+ "cache_write_input_tokens": 0,
153
+ "cached_input_tokens": 24960,
154
+ "input_tokens": 27403,
155
+ "output_tokens": 385,
156
+ "reasoning_output_tokens": 0
157
+ },
158
+ {
159
+ "cache_write_input_tokens": 0,
160
+ "cached_input_tokens": 27264,
161
+ "input_tokens": 27811,
162
+ "output_tokens": 39,
163
+ "reasoning_output_tokens": 0
164
+ },
165
+ {
166
+ "cache_write_input_tokens": 0,
167
+ "cached_input_tokens": 27648,
168
+ "input_tokens": 28265,
169
+ "output_tokens": 61,
170
+ "reasoning_output_tokens": 22
171
+ },
172
+ {
173
+ "cache_write_input_tokens": 0,
174
+ "cached_input_tokens": 28032,
175
+ "input_tokens": 28738,
176
+ "output_tokens": 38,
177
+ "reasoning_output_tokens": 0
178
+ },
179
+ {
180
+ "cache_write_input_tokens": 0,
181
+ "cached_input_tokens": 28544,
182
+ "input_tokens": 29189,
183
+ "output_tokens": 38,
184
+ "reasoning_output_tokens": 0
185
+ },
186
+ {
187
+ "cache_write_input_tokens": 0,
188
+ "cached_input_tokens": 28928,
189
+ "input_tokens": 29640,
190
+ "output_tokens": 49,
191
+ "reasoning_output_tokens": 9
192
+ },
193
+ {
194
+ "cache_write_input_tokens": 0,
195
+ "cached_input_tokens": 29568,
196
+ "input_tokens": 30101,
197
+ "output_tokens": 37,
198
+ "reasoning_output_tokens": 0
199
+ },
200
+ {
201
+ "cache_write_input_tokens": 0,
202
+ "cached_input_tokens": 29952,
203
+ "input_tokens": 30552,
204
+ "output_tokens": 37,
205
+ "reasoning_output_tokens": 0
206
+ },
207
+ {
208
+ "cache_write_input_tokens": 0,
209
+ "cached_input_tokens": 30464,
210
+ "input_tokens": 31003,
211
+ "output_tokens": 38,
212
+ "reasoning_output_tokens": 0
213
+ },
214
+ {
215
+ "cache_write_input_tokens": 0,
216
+ "cached_input_tokens": 30848,
217
+ "input_tokens": 31456,
218
+ "output_tokens": 39,
219
+ "reasoning_output_tokens": 0
220
+ },
221
+ {
222
+ "cache_write_input_tokens": 0,
223
+ "cached_input_tokens": 31360,
224
+ "input_tokens": 31913,
225
+ "output_tokens": 39,
226
+ "reasoning_output_tokens": 0
227
+ },
228
+ {
229
+ "cache_write_input_tokens": 0,
230
+ "cached_input_tokens": 31744,
231
+ "input_tokens": 32368,
232
+ "output_tokens": 39,
233
+ "reasoning_output_tokens": 0
234
+ },
235
+ {
236
+ "cache_write_input_tokens": 0,
237
+ "cached_input_tokens": 32128,
238
+ "input_tokens": 32821,
239
+ "output_tokens": 38,
240
+ "reasoning_output_tokens": 0
241
+ },
242
+ {
243
+ "cache_write_input_tokens": 0,
244
+ "cached_input_tokens": 32640,
245
+ "input_tokens": 33269,
246
+ "output_tokens": 414,
247
+ "reasoning_output_tokens": 35
248
+ },
249
+ {
250
+ "cache_write_input_tokens": 0,
251
+ "cached_input_tokens": 33152,
252
+ "input_tokens": 33712,
253
+ "output_tokens": 72,
254
+ "reasoning_output_tokens": 0
255
+ },
256
+ {
257
+ "cache_write_input_tokens": 0,
258
+ "cached_input_tokens": 33664,
259
+ "input_tokens": 34232,
260
+ "output_tokens": 44,
261
+ "reasoning_output_tokens": 0
262
+ },
263
+ {
264
+ "cache_write_input_tokens": 0,
265
+ "cached_input_tokens": 34176,
266
+ "input_tokens": 34724,
267
+ "output_tokens": 43,
268
+ "reasoning_output_tokens": 0
269
+ },
270
+ {
271
+ "cache_write_input_tokens": 0,
272
+ "cached_input_tokens": 34560,
273
+ "input_tokens": 35218,
274
+ "output_tokens": 45,
275
+ "reasoning_output_tokens": 0
276
+ },
277
+ {
278
+ "cache_write_input_tokens": 0,
279
+ "cached_input_tokens": 35072,
280
+ "input_tokens": 35712,
281
+ "output_tokens": 59,
282
+ "reasoning_output_tokens": 12
283
+ },
284
+ {
285
+ "cache_write_input_tokens": 0,
286
+ "cached_input_tokens": 35584,
287
+ "input_tokens": 36237,
288
+ "output_tokens": 45,
289
+ "reasoning_output_tokens": 0
290
+ },
291
+ {
292
+ "cache_write_input_tokens": 0,
293
+ "cached_input_tokens": 36096,
294
+ "input_tokens": 36747,
295
+ "output_tokens": 67,
296
+ "reasoning_output_tokens": 20
297
+ },
298
+ {
299
+ "cache_write_input_tokens": 0,
300
+ "cached_input_tokens": 36608,
301
+ "input_tokens": 37276,
302
+ "output_tokens": 59,
303
+ "reasoning_output_tokens": 20
304
+ },
305
+ {
306
+ "cache_write_input_tokens": 0,
307
+ "cached_input_tokens": 37120,
308
+ "input_tokens": 37660,
309
+ "output_tokens": 65,
310
+ "reasoning_output_tokens": 21
311
+ },
312
+ {
313
+ "cache_write_input_tokens": 0,
314
+ "cached_input_tokens": 37504,
315
+ "input_tokens": 37935,
316
+ "output_tokens": 129,
317
+ "reasoning_output_tokens": 15
318
+ },
319
+ {
320
+ "cache_write_input_tokens": 0,
321
+ "cached_input_tokens": 37888,
322
+ "input_tokens": 38650,
323
+ "output_tokens": 33,
324
+ "reasoning_output_tokens": 0
325
+ },
326
+ {
327
+ "cache_write_input_tokens": 0,
328
+ "cached_input_tokens": 38528,
329
+ "input_tokens": 38788,
330
+ "output_tokens": 105,
331
+ "reasoning_output_tokens": 29
332
+ },
333
+ {
334
+ "cache_write_input_tokens": 0,
335
+ "cached_input_tokens": 38656,
336
+ "input_tokens": 39125,
337
+ "output_tokens": 220,
338
+ "reasoning_output_tokens": 0
339
+ },
340
+ {
341
+ "cache_write_input_tokens": 0,
342
+ "cached_input_tokens": 39040,
343
+ "input_tokens": 39697,
344
+ "output_tokens": 558,
345
+ "reasoning_output_tokens": 60
346
+ },
347
+ {
348
+ "cache_write_input_tokens": 0,
349
+ "cached_input_tokens": 39552,
350
+ "input_tokens": 40278,
351
+ "output_tokens": 63,
352
+ "reasoning_output_tokens": 0
353
+ },
354
+ {
355
+ "cache_write_input_tokens": 0,
356
+ "cached_input_tokens": 40192,
357
+ "input_tokens": 40564,
358
+ "output_tokens": 53,
359
+ "reasoning_output_tokens": 0
360
+ },
361
+ {
362
+ "cache_write_input_tokens": 0,
363
+ "cached_input_tokens": 40448,
364
+ "input_tokens": 40840,
365
+ "output_tokens": 61,
366
+ "reasoning_output_tokens": 0
367
+ },
368
+ {
369
+ "cache_write_input_tokens": 0,
370
+ "cached_input_tokens": 40704,
371
+ "input_tokens": 41123,
372
+ "output_tokens": 45,
373
+ "reasoning_output_tokens": 0
374
+ },
375
+ {
376
+ "cache_write_input_tokens": 0,
377
+ "cached_input_tokens": 40960,
378
+ "input_tokens": 41615,
379
+ "output_tokens": 58,
380
+ "reasoning_output_tokens": 12
381
+ },
382
+ {
383
+ "cache_write_input_tokens": 0,
384
+ "cached_input_tokens": 41472,
385
+ "input_tokens": 42118,
386
+ "output_tokens": 46,
387
+ "reasoning_output_tokens": 0
388
+ },
389
+ {
390
+ "cache_write_input_tokens": 0,
391
+ "cached_input_tokens": 41984,
392
+ "input_tokens": 42611,
393
+ "output_tokens": 45,
394
+ "reasoning_output_tokens": 0
395
+ },
396
+ {
397
+ "cache_write_input_tokens": 0,
398
+ "cached_input_tokens": 42368,
399
+ "input_tokens": 43105,
400
+ "output_tokens": 46,
401
+ "reasoning_output_tokens": 0
402
+ },
403
+ {
404
+ "cache_write_input_tokens": 0,
405
+ "cached_input_tokens": 43008,
406
+ "input_tokens": 43598,
407
+ "output_tokens": 81,
408
+ "reasoning_output_tokens": 10
409
+ },
410
+ {
411
+ "cache_write_input_tokens": 0,
412
+ "cached_input_tokens": 43520,
413
+ "input_tokens": 44118,
414
+ "output_tokens": 43,
415
+ "reasoning_output_tokens": 0
416
+ },
417
+ {
418
+ "cache_write_input_tokens": 0,
419
+ "cached_input_tokens": 44032,
420
+ "input_tokens": 44607,
421
+ "output_tokens": 46,
422
+ "reasoning_output_tokens": 0
423
+ },
424
+ {
425
+ "cache_write_input_tokens": 0,
426
+ "cached_input_tokens": 44416,
427
+ "input_tokens": 45102,
428
+ "output_tokens": 79,
429
+ "reasoning_output_tokens": 14
430
+ },
431
+ {
432
+ "cache_write_input_tokens": 0,
433
+ "cached_input_tokens": 44928,
434
+ "input_tokens": 45404,
435
+ "output_tokens": 46,
436
+ "reasoning_output_tokens": 0
437
+ },
438
+ {
439
+ "cache_write_input_tokens": 0,
440
+ "cached_input_tokens": 45184,
441
+ "input_tokens": 45896,
442
+ "output_tokens": 47,
443
+ "reasoning_output_tokens": 0
444
+ },
445
+ {
446
+ "cache_write_input_tokens": 0,
447
+ "cached_input_tokens": 45696,
448
+ "input_tokens": 46388,
449
+ "output_tokens": 45,
450
+ "reasoning_output_tokens": 0
451
+ },
452
+ {
453
+ "cache_write_input_tokens": 0,
454
+ "cached_input_tokens": 46208,
455
+ "input_tokens": 46878,
456
+ "output_tokens": 43,
457
+ "reasoning_output_tokens": 0
458
+ },
459
+ {
460
+ "cache_write_input_tokens": 0,
461
+ "cached_input_tokens": 46720,
462
+ "input_tokens": 47354,
463
+ "output_tokens": 44,
464
+ "reasoning_output_tokens": 0
465
+ },
466
+ {
467
+ "cache_write_input_tokens": 0,
468
+ "cached_input_tokens": 47232,
469
+ "input_tokens": 47529,
470
+ "output_tokens": 28,
471
+ "reasoning_output_tokens": 0
472
+ },
473
+ {
474
+ "cache_write_input_tokens": 0,
475
+ "cached_input_tokens": 47360,
476
+ "input_tokens": 47619,
477
+ "output_tokens": 49,
478
+ "reasoning_output_tokens": 0
479
+ },
480
+ {
481
+ "cache_write_input_tokens": 0,
482
+ "cached_input_tokens": 47488,
483
+ "input_tokens": 48050,
484
+ "output_tokens": 531,
485
+ "reasoning_output_tokens": 39
486
+ },
487
+ {
488
+ "cache_write_input_tokens": 0,
489
+ "cached_input_tokens": 47872,
490
+ "input_tokens": 48610,
491
+ "output_tokens": 39,
492
+ "reasoning_output_tokens": 0
493
+ }
494
+ ],
495
+ "usage_source": "provider-raw"
496
+ },
497
+ "call_activity": {
498
+ "schema": "rlebench/call-activity/1",
499
+ "source": "Kinex native trace",
500
+ "model_tool_calls": 69,
501
+ "model_tool_calls_by_name": {
502
+ "ls": 3,
503
+ "read": 15,
504
+ "find": 1,
505
+ "exec": 40,
506
+ "devices": 3,
507
+ "device": 1,
508
+ "write": 3,
509
+ "edit": 1,
510
+ "relative_pose": 1,
511
+ "device_stop": 1
512
+ },
513
+ "python_instrumented_model_tool_calls": null,
514
+ "python_tool_invocations": null,
515
+ "nested_python_tool_invocations": null,
516
+ "python_tool_invocations_by_origin": null,
517
+ "python_device_rpc_attempts": null,
518
+ "python_device_rpc_attempts_by_action": null,
519
+ "python_device_rpc_errors": null,
520
+ "python_scope": "Recorded instrumented Python invocations only; other paths and automatic fatal stops are not counted."
521
+ },
522
+ "model_response_count": 60,
523
+ "usage_source": "Harbor session metrics",
524
+ "execution": {
525
+ "status": "finished",
526
+ "reason": null
527
+ },
528
+ "native": {
529
+ "evidence_valid": true,
530
+ "success": true,
531
+ "termination": "success",
532
+ "steps": 1277
533
+ },
534
+ "failure_attribution": {
535
+ "classification": "native_success",
536
+ "physical_cause": null,
537
+ "physical_cause_evidence": null
538
+ },
539
+ "exception": null
540
+ }
experiments/robotwin-native/kinex_policy-3/episode.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "episode": 1,
3
+ "initialization": {
4
+ "expert_plan_success": false,
5
+ "requested_seed": 0
6
+ },
7
+ "native_reward": 1.0,
8
+ "planned_episodes": 1,
9
+ "reason": "success",
10
+ "simulation_time_s": 51.08,
11
+ "step_limit": 1700,
12
+ "steps": 1277,
13
+ "success": true,
14
+ "valid": true
15
+ }
experiments/robotwin-native/kinex_policy-3/evidence.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c3183b3564c73f7a6d86b9081ad762d7ead9377d1478f2722c3ce5eb13fac3c
3
+ size 15425702
experiments/robotwin-native/kinex_policy-3/instruction.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "binding_paths": {},
3
+ "bindings": {},
4
+ "context": {},
5
+ "instruction": "use arms to grab the bottles and put them into the dustbin to the left of the table",
6
+ "native_instruction": "",
7
+ "policy": "original_native",
8
+ "prompt": "# Task: robotwin/put-bottles-dustbin-aloha-agilex\n\nuse arms to grab the bottles and put them into the dustbin to the left of the table\n\n## Work to do\n\nSolve the task efficiently using the running rosbridge environment. Follow the\ntask instruction above, including its completion requirements. Read `harness.md`\nand the device profiles for public interfaces, units and calibrated frames.\nReuse your saved tools, skills, and memos.\n\n## Limits\n\nYou have 1700 native action frames. The execution mode is `stepped`: physics is\nfrozen while idle. Use `world.wait` with `duration_s` to advance the native clock.\nJoint targets are ROS JointTrajectory messages; gripper targets use the units\ndeclared in their profiles. Reads and command publication do not advance physics.\nWaits are sequential; concurrent waits are rejected. A command or elapsed time\ndoes not prove arrival; inspect measured feedback. Device stop holds the device\nwithout ending the episode. The safety timeout is 28800 seconds.\nDo not launch or reset a simulation.\n\n## Completion\n\nReturn when finished. The owner finalizes the episode and reads its native\nverdict. An interrupted or rejected wait does not establish task success. The\nnative step limit stops physics even if your requested wait is longer.\n\n## Saved resources\n\nMaintain reusable code in `tools/`, skills in `skills/`, and concise working notes\nin `memos/`. Improve existing files in place as you learn, replacing outdated\ncontent. Create a new file only when no existing resource fits; do not create\nper-attempt reflections or numbered copies. These files persist across fresh\nconversations. Update them before finishing your response.\n\n## Robot and device access\n\nUse Kinex's ROS device profiles in `.kinex/devices/` and the robot\nspecification in `.kinex/workspace.yaml`. The adapter refreshes them each episode.\nUse `kinex.communication` for typed topic publication, subscriptions and service\nrequests. Inspect operation schemas before acting. Load the provided devices;\nthey borrow the running simulator. Stop holds motion; it does not reset the world.\nImprove existing tools, skills, and memos in place.\n\n",
9
+ "prompt_template": "<!-- rlebench:resolved-episode-instruction -->\n# Task: robotwin/put-bottles-dustbin-aloha-agilex\n\nuse arms to grab the bottles and put them into the dustbin to the left of the table\n\n## Work to do\n\nSolve the task efficiently using the running rosbridge environment. Follow the\ntask instruction above, including its completion requirements. Read `harness.md`\nand the device profiles for public interfaces, units and calibrated frames.\nReuse your saved tools, skills, and memos.\n\n## Limits\n\nYou have 1700 native action frames. The execution mode is `stepped`: physics is\nfrozen while idle. Use `world.wait` with `duration_s` to advance the native clock.\nJoint targets are ROS JointTrajectory messages; gripper targets use the units\ndeclared in their profiles. Reads and command publication do not advance physics.\nWaits are sequential; concurrent waits are rejected. A command or elapsed time\ndoes not prove arrival; inspect measured feedback. Device stop holds the device\nwithout ending the episode. The safety timeout is 28800 seconds.\nDo not launch or reset a simulation.\n\n## Completion\n\nReturn when finished. The owner finalizes the episode and reads its native\nverdict. An interrupted or rejected wait does not establish task success. The\nnative step limit stops physics even if your requested wait is longer.\n\n## Saved resources\n\nMaintain reusable code in `tools/`, skills in `skills/`, and concise working notes\nin `memos/`. Improve existing files in place as you learn, replacing outdated\ncontent. Create a new file only when no existing resource fits; do not create\nper-attempt reflections or numbered copies. These files persist across fresh\nconversations. Update them before finishing your response.\n\n## Robot and device access\n\nUse Kinex's ROS device profiles in `.kinex/devices/` and the robot\nspecification in `.kinex/workspace.yaml`. The adapter refreshes them each episode.\nUse `kinex.communication` for typed topic publication, subscriptions and service\nrequests. Inspect operation schemas before acting. Load the provided devices;\nthey borrow the running simulator. Stop holds motion; it does not reset the world.\nImprove existing tools, skills, and memos in place.\n\n",
10
+ "schema": "rlebench/episode-instruction/3",
11
+ "source": "catalog",
12
+ "task_id": "robotwin/put-bottles-dustbin-aloha-agilex",
13
+ "template": "use arms to grab the bottles and put them into the dustbin to the left of the table"
14
+ }
experiments/robotwin-native/kinex_policy-3/media.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "native_frames": 512,
3
+ "published_frames": 512,
4
+ "pts_verified": true,
5
+ "source_start_pts": 0.2,
6
+ "published_start_pts": 0.0,
7
+ "pts_offset": -0.2,
8
+ "terminal_frame_retained": true,
9
+ "views": [
10
+ "third_person",
11
+ "left_wrist",
12
+ "right_wrist"
13
+ ],
14
+ "fps": "10/1",
15
+ "native_recording": {
16
+ "schema": "rosbridge/recording/1",
17
+ "run_id": "88c15864ae34",
18
+ "task": "robotwin/put-bottles-dustbin-aloha-agilex",
19
+ "state": "complete",
20
+ "views": {
21
+ "third_person": {
22
+ "received": 512,
23
+ "dropped": 0,
24
+ "segments": [
25
+ {
26
+ "file": "third_person.mp4",
27
+ "frames": 512,
28
+ "width": 960,
29
+ "height": 720,
30
+ "fps": 10.0,
31
+ "first_stamp_ns": 0,
32
+ "last_stamp_ns": 51080000000
33
+ }
34
+ ],
35
+ "error": null
36
+ },
37
+ "left_wrist": {
38
+ "received": 512,
39
+ "dropped": 0,
40
+ "segments": [
41
+ {
42
+ "file": "left_wrist.mp4",
43
+ "frames": 512,
44
+ "width": 960,
45
+ "height": 720,
46
+ "fps": 10.0,
47
+ "first_stamp_ns": 0,
48
+ "last_stamp_ns": 51080000000
49
+ }
50
+ ],
51
+ "error": null
52
+ },
53
+ "right_wrist": {
54
+ "received": 512,
55
+ "dropped": 0,
56
+ "segments": [
57
+ {
58
+ "file": "right_wrist.mp4",
59
+ "frames": 512,
60
+ "width": 960,
61
+ "height": 720,
62
+ "fps": 10.0,
63
+ "first_stamp_ns": 0,
64
+ "last_stamp_ns": 51080000000
65
+ }
66
+ ],
67
+ "error": null
68
+ }
69
+ },
70
+ "type": "mp4",
71
+ "spectator_views": [
72
+ "third_person",
73
+ "left_wrist",
74
+ "right_wrist"
75
+ ],
76
+ "fps": 10.0,
77
+ "topics": [],
78
+ "error": null,
79
+ "bag_messages": 0
80
+ },
81
+ "all_frame_visual_checks": {
82
+ "third_person": {
83
+ "frames_compared": 512,
84
+ "mean_ssim": 0.998352236328125,
85
+ "minimum_ssim": 0.996492,
86
+ "label_area_excluded": true
87
+ },
88
+ "left_wrist": {
89
+ "frames_compared": 512,
90
+ "mean_ssim": 0.99676091015625,
91
+ "minimum_ssim": 0.991867,
92
+ "label_area_excluded": true
93
+ },
94
+ "right_wrist": {
95
+ "frames_compared": 512,
96
+ "mean_ssim": 0.99658952734375,
97
+ "minimum_ssim": 0.993767,
98
+ "label_area_excluded": true
99
+ }
100
+ }
101
+ }
experiments/robotwin-native/kinex_policy-3/observations/0b176341-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/13ff48a6-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/173a92e8-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/1d570650-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/265cda94-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/27bf832e-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/2cb2e601-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/2ced7a9d-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/36fe7431-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/36fe7431-2.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/40c24470-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/41da9eb5-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/44263e03-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/6d700e8b-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/6f0ea5a7-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/6f40ac77-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/7ad93479-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/7d1d9cc1-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/7d40da85-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/806acf8c-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/81da75c1-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/82bb4d8a-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/92a89f2b-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/935284a6-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/9ba4e93f-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/9e391af7-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/a119dcde-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/a39bc7a3-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/a85fdbef-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/a8978258-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/b03aa410-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/b0b9d8e3-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/b2b427a8-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/ba87585e-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/c5bb466a-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/c5bb466a-2.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/d2b3d472-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/d4a17c8a-1.png ADDED
experiments/robotwin-native/kinex_policy-3/observations/e8c5e432-1.png ADDED