yzzhao commited on
Commit
518ecbf
·
verified ·
1 Parent(s): 8e48ad6

Publish eight native episodes with policy failure and recovery traces

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 +2 -0
  2. app.js +1 -1
  3. data.json +29 -3
  4. experiments/robotwin-native/kinex-3/artifacts.json +109 -0
  5. experiments/robotwin-native/kinex-3/combined.mp4 +3 -0
  6. experiments/robotwin-native/kinex-3/detail.json +1305 -0
  7. experiments/robotwin-native/kinex-3/effective-model.json +16 -0
  8. experiments/robotwin-native/kinex-3/episode-report.json +462 -0
  9. experiments/robotwin-native/kinex-3/episode.json +15 -0
  10. experiments/robotwin-native/kinex-3/evidence.tar.gz +3 -0
  11. experiments/robotwin-native/kinex-3/instruction.json +14 -0
  12. experiments/robotwin-native/kinex-3/media.json +101 -0
  13. experiments/robotwin-native/kinex-3/observations/1c00b1c6-1.png +0 -0
  14. experiments/robotwin-native/kinex-3/observations/1c00b1c6-2.png +0 -0
  15. experiments/robotwin-native/kinex-3/observations/25fb6783-1.png +0 -0
  16. experiments/robotwin-native/kinex-3/observations/2e34d545-1.png +0 -0
  17. experiments/robotwin-native/kinex-3/observations/38f9e622-1.png +0 -0
  18. experiments/robotwin-native/kinex-3/observations/38f9e622-2.png +0 -0
  19. experiments/robotwin-native/kinex-3/observations/40cd7359-1.png +0 -0
  20. experiments/robotwin-native/kinex-3/observations/4a031853-1.png +0 -0
  21. experiments/robotwin-native/kinex-3/observations/4a031853-2.png +0 -0
  22. experiments/robotwin-native/kinex-3/observations/5306339c-1.png +0 -0
  23. experiments/robotwin-native/kinex-3/observations/5306339c-2.png +0 -0
  24. experiments/robotwin-native/kinex-3/observations/59f4e1b6-1.png +0 -0
  25. experiments/robotwin-native/kinex-3/observations/7af7cb21-1.png +0 -0
  26. experiments/robotwin-native/kinex-3/observations/82847509-1.png +0 -0
  27. experiments/robotwin-native/kinex-3/observations/8c8cb210-1.png +0 -0
  28. experiments/robotwin-native/kinex-3/observations/8c8cb210-2.png +0 -0
  29. experiments/robotwin-native/kinex-3/observations/9deb50d9-1.png +0 -0
  30. experiments/robotwin-native/kinex-3/observations/9deb50d9-2.png +0 -0
  31. experiments/robotwin-native/kinex-3/observations/a2501cc5-1.png +0 -0
  32. experiments/robotwin-native/kinex-3/observations/a57d6a5b-1.png +0 -0
  33. experiments/robotwin-native/kinex-3/observations/b27b5e13-1.png +0 -0
  34. experiments/robotwin-native/kinex-3/observations/bd40c7f2-1.png +0 -0
  35. experiments/robotwin-native/kinex-3/observations/bd40c7f2-2.png +0 -0
  36. experiments/robotwin-native/kinex-3/observations/e0aba462-1.png +0 -0
  37. experiments/robotwin-native/kinex-3/observations/e8dcb044-1.png +0 -0
  38. experiments/robotwin-native/kinex-3/observations/e8dcb044-2.png +0 -0
  39. experiments/robotwin-native/kinex-3/observations/ed3b0775-1.png +0 -0
  40. experiments/robotwin-native/kinex-3/observations/ed3b0775-2.png +0 -0
  41. experiments/robotwin-native/kinex-3/observations/eda38a00-1.png +0 -0
  42. experiments/robotwin-native/kinex-3/observations/eda38a00-2.png +0 -0
  43. experiments/robotwin-native/kinex-3/observations/ff2d32db-1.png +0 -0
  44. experiments/robotwin-native/kinex-3/policy-calls.json +1 -0
  45. experiments/robotwin-native/kinex-3/poster.jpg +0 -0
  46. experiments/robotwin-native/kinex-3/provider-requests.jsonl +0 -0
  47. experiments/robotwin-native/kinex-3/session.json +0 -0
  48. experiments/robotwin-native/kinex-3/session.jsonl +0 -0
  49. experiments/robotwin-native/kinex-3/terminal.jpg +0 -0
  50. experiments/robotwin-native/kinex-3/tool-calls.json +0 -0
.gitattributes CHANGED
@@ -60,3 +60,5 @@ experiments/robotwin-native/kinex_policy-1/combined.mp4 filter=lfs diff=lfs merg
60
  experiments/robotwin-native/standalone-2/combined.mp4 filter=lfs diff=lfs merge=lfs -text
61
  experiments/robotwin-native/kinex-2/combined.mp4 filter=lfs diff=lfs merge=lfs -text
62
  experiments/robotwin-native/standalone-3/combined.mp4 filter=lfs diff=lfs merge=lfs -text
 
 
 
60
  experiments/robotwin-native/standalone-2/combined.mp4 filter=lfs diff=lfs merge=lfs -text
61
  experiments/robotwin-native/kinex-2/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
app.js CHANGED
@@ -17,6 +17,6 @@ else if(tab==='artifacts')target.innerHTML=`<div class="panel files"><h2>Recorde
17
  else if(tab==='session'){const v=await load(e.path+'/session.json');target.innerHTML=`<p class="muted">${num(v.length)} messages · <a href="${e.path}/session.json">Download full session</a></p>${v.map(m=>`<article class="message"><b>${esc(m.role)} ${esc(m.name||'')}</b><div class="text">${esc(m.text)}</div><div class="images">${(m.images||[]).map(x=>`<a href="${esc(x)}" target="_blank"><img loading="lazy" src="${esc(x)}" alt="Agent observation"></a>`).join('')}</div>${m.data?`<details><summary>Recorded payload</summary><pre>${esc(JSON.stringify(m.data,null,2))}</pre></details>`:''}</article>`).join('')||'<div class="empty">Standalone evaluation has no planning-agent session.</div>'}`;}
18
  else if(tab==='tools'){const v=await load(e.path+'/tool-calls.json');target.innerHTML=`<p class="muted">${num(v.length)} calls · <a href="${e.path}/tool-calls.json">Download tool calls</a></p>${v.map((m,i)=>`<details><summary>${i+1}. ${esc(m.name)} <span class="muted">${esc(m.summary||'')}</span></summary><h3>Input</h3><pre>${esc(JSON.stringify(m.input,null,2))}</pre><h3>Output</h3><pre>${esc(typeof m.output==='string'?m.output:JSON.stringify(m.output,null,2))}</pre></details>`).join('')||'<div class="empty">No agent tool calls</div>'}`;}
19
  else if(tab==='policy'){const v=await load(e.path+'/policy-calls.json');target.innerHTML=`<p class="muted">${num(v.length)} inferences · <a href="${e.path}/policy-calls.json">Download all input / output records</a></p>${v.map((m,i)=>`<details><summary>${i+1}. ${esc(m.task)} <span class="muted">${m.inference_s?.toFixed(2)||'—'} s</span></summary><div class="images">${(m.images||[]).map(x=>`<a href="${esc(x)}" target="_blank"><img loading="lazy" src="${esc(x)}" alt="Policy input observation"></a>`).join('')}</div><h3>Predicted actions and interface conversion</h3><pre>${esc(JSON.stringify(m.output,null,2))}</pre></details>`).join('')||'<div class="empty">No VLA inference recorded</div>'}`;}}
20
- function protocol(){$('#main').innerHTML=`<div class="eyebrow">Experiment protocol</div><h1>One task. Three control choices.</h1><div class="panel"><h2>Execution</h2><table><tbody>${[['Suite','RLE-Bench task03 · RoboTwin 2.0 · ALOHA AgileX'],['VLA','StarVLA/Qwen3-VL-OFT-Robotwin2 · steps 40,000'],['Agent','Kinex · GPT-6 Astra · high'],['Policy interface','Native run_policy · one inference per call · agent owns the outer loop'],['Control','25 Hz native joint / gripper frames; original task-specific limits'],['Selection','First three task03 tasks, seed 0; frozen before outcome inspection'],['Scope','Frozen-policy behavior only; no encoder or progress monitor']].map(([k,v])=>`<tr><th>${k}</th><td>${v}</td></tr>`).join('')}</tbody></table></div><p class="muted">This fixed-rate device protocol differs from upstream VLA trajectory execution. Published StarVLA benchmark scores are not a direct comparison. RoboTwin 2.0-Plus is a separate suite and is not included here.</p><a href="sources.json">Source and model identities ↗</a>`}
21
  async function render(){try{const parts=location.hash.slice(1).split('/');if(parts[0]==='episode')await detail(parts[1],parts[2],parts[3]||'overview');else if(parts[0]==='protocol')protocol();else overview(parts[0]==='tasks')}catch(e){$('#main').innerHTML=`<div class="panel"><h1>Could not load this record</h1><p>${esc(e.message)}</p><a href="#overview">Return to overview</a></div>`}}
22
  fetch('data.json').then(r=>r.json()).then(x=>{data=x;window.addEventListener('hashchange',render);render()}).catch(e=>{$('#main').textContent=e.message});
 
17
  else if(tab==='session'){const v=await load(e.path+'/session.json');target.innerHTML=`<p class="muted">${num(v.length)} messages · <a href="${e.path}/session.json">Download full session</a></p>${v.map(m=>`<article class="message"><b>${esc(m.role)} ${esc(m.name||'')}</b><div class="text">${esc(m.text)}</div><div class="images">${(m.images||[]).map(x=>`<a href="${esc(x)}" target="_blank"><img loading="lazy" src="${esc(x)}" alt="Agent observation"></a>`).join('')}</div>${m.data?`<details><summary>Recorded payload</summary><pre>${esc(JSON.stringify(m.data,null,2))}</pre></details>`:''}</article>`).join('')||'<div class="empty">Standalone evaluation has no planning-agent session.</div>'}`;}
18
  else if(tab==='tools'){const v=await load(e.path+'/tool-calls.json');target.innerHTML=`<p class="muted">${num(v.length)} calls · <a href="${e.path}/tool-calls.json">Download tool calls</a></p>${v.map((m,i)=>`<details><summary>${i+1}. ${esc(m.name)} <span class="muted">${esc(m.summary||'')}</span></summary><h3>Input</h3><pre>${esc(JSON.stringify(m.input,null,2))}</pre><h3>Output</h3><pre>${esc(typeof m.output==='string'?m.output:JSON.stringify(m.output,null,2))}</pre></details>`).join('')||'<div class="empty">No agent tool calls</div>'}`;}
19
  else if(tab==='policy'){const v=await load(e.path+'/policy-calls.json');target.innerHTML=`<p class="muted">${num(v.length)} inferences · <a href="${e.path}/policy-calls.json">Download all input / output records</a></p>${v.map((m,i)=>`<details><summary>${i+1}. ${esc(m.task)} <span class="muted">${m.inference_s?.toFixed(2)||'—'} s</span></summary><div class="images">${(m.images||[]).map(x=>`<a href="${esc(x)}" target="_blank"><img loading="lazy" src="${esc(x)}" alt="Policy input observation"></a>`).join('')}</div><h3>Predicted actions and interface conversion</h3><pre>${esc(JSON.stringify(m.output,null,2))}</pre></details>`).join('')||'<div class="empty">No VLA inference recorded</div>'}`;}}
20
+ function protocol(){$('#main').innerHTML=`<div class="eyebrow">Experiment protocol</div><h1>One task. Three control choices.</h1><div class="panel"><h2>Execution</h2><table><tbody>${[['Suite','RLE-Bench task03 · RoboTwin 2.0 · ALOHA AgileX'],['VLA','StarVLA/Qwen3-VL-OFT-Robotwin2 · steps 40,000'],['Agent','Kinex · GPT-6 Astra · high'],['Policy interface','Native run_policy · one inference per call · agent owns the outer loop'],['Inputs','VLA: three RGB views and language. Kinex: public RGB-D, joint state, calibration and native tools.'],['Control','25 Hz native joint / gripper frames; original task-specific limits'],['Selection','First three task03 tasks, seed 0; frozen before outcome inspection'],['Scope','Frozen-policy behavior only; no encoder or progress monitor']].map(([k,v])=>`<tr><th>${k}</th><td>${v}</td></tr>`).join('')}</tbody></table></div><p class="muted">This fixed-rate device protocol differs from upstream VLA trajectory execution. Published StarVLA benchmark scores are not a direct comparison. RoboTwin 2.0-Plus is a separate suite and is not included here.</p><a href="sources.json">Source and model identities ↗</a>`}
21
  async function render(){try{const parts=location.hash.slice(1).split('/');if(parts[0]==='episode')await detail(parts[1],parts[2],parts[3]||'overview');else if(parts[0]==='protocol')protocol();else overview(parts[0]==='tasks')}catch(e){$('#main').innerHTML=`<div class="panel"><h1>Could not load this record</h1><p>${esc(e.message)}</p><a href="#overview">Return to overview</a></div>`}}
22
  fetch('data.json').then(r=>r.json()).then(x=>{data=x;window.addEventListener('hashchange',render);render()}).catch(e=>{$('#main').textContent=e.message});
data.json CHANGED
@@ -139,7 +139,20 @@
139
  "id": "kinex_policy-2",
140
  "task": "2",
141
  "condition": "kinex_policy",
142
- "status": "running"
 
 
 
 
 
 
 
 
 
 
 
 
 
143
  },
144
  {
145
  "id": "standalone-3",
@@ -164,13 +177,26 @@
164
  "id": "kinex-3",
165
  "task": "3",
166
  "condition": "kinex",
167
- "status": "running"
 
 
 
 
 
 
 
 
 
 
 
 
 
168
  },
169
  {
170
  "id": "kinex_policy-3",
171
  "task": "3",
172
  "condition": "kinex_policy",
173
- "status": "pending"
174
  }
175
  ]
176
  }
 
139
  "id": "kinex_policy-2",
140
  "task": "2",
141
  "condition": "kinex_policy",
142
+ "status": "complete",
143
+ "success": false,
144
+ "valid": true,
145
+ "control_steps": 498,
146
+ "tool_calls": 72,
147
+ "vla_calls": 15,
148
+ "input_tokens": 2354746,
149
+ "output_tokens": 9984,
150
+ "cached_input_tokens": 2249856,
151
+ "api_cost_usd": 3.797956,
152
+ "path": "experiments/robotwin-native/kinex_policy-2",
153
+ "video": "experiments/robotwin-native/kinex_policy-2/combined.mp4",
154
+ "poster": "experiments/robotwin-native/kinex_policy-2/poster.jpg",
155
+ "wall_time_s": 817.625303
156
  },
157
  {
158
  "id": "standalone-3",
 
177
  "id": "kinex-3",
178
  "task": "3",
179
  "condition": "kinex",
180
+ "status": "complete",
181
+ "success": true,
182
+ "valid": true,
183
+ "control_steps": 1354,
184
+ "tool_calls": 59,
185
+ "vla_calls": 0,
186
+ "input_tokens": 1644067,
187
+ "output_tokens": 10285,
188
+ "cached_input_tokens": 1598336,
189
+ "api_cost_usd": 2.569896,
190
+ "path": "experiments/robotwin-native/kinex-3",
191
+ "video": "experiments/robotwin-native/kinex-3/combined.mp4",
192
+ "poster": "experiments/robotwin-native/kinex-3/poster.jpg",
193
+ "wall_time_s": 756.194745
194
  },
195
  {
196
  "id": "kinex_policy-3",
197
  "task": "3",
198
  "condition": "kinex_policy",
199
+ "status": "running"
200
  }
201
  ]
202
  }
experiments/robotwin-native/kinex-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": "c3ff59536ab0",
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": 543,
31
+ "segments": [
32
+ {
33
+ "file": "left_wrist.mp4",
34
+ "first_stamp_ns": 0,
35
+ "fps": 10.0,
36
+ "frames": 543,
37
+ "height": 720,
38
+ "last_stamp_ns": 54160000000,
39
+ "width": 960
40
+ }
41
+ ]
42
+ },
43
+ "right_wrist": {
44
+ "dropped": 0,
45
+ "error": null,
46
+ "received": 543,
47
+ "segments": [
48
+ {
49
+ "file": "right_wrist.mp4",
50
+ "first_stamp_ns": 0,
51
+ "fps": 10.0,
52
+ "frames": 543,
53
+ "height": 720,
54
+ "last_stamp_ns": 54160000000,
55
+ "width": 960
56
+ }
57
+ ]
58
+ },
59
+ "third_person": {
60
+ "dropped": 0,
61
+ "error": null,
62
+ "received": 543,
63
+ "segments": [
64
+ {
65
+ "file": "third_person.mp4",
66
+ "first_stamp_ns": 0,
67
+ "fps": 10.0,
68
+ "frames": 543,
69
+ "height": 720,
70
+ "last_stamp_ns": 54160000000,
71
+ "width": 960
72
+ }
73
+ ]
74
+ }
75
+ }
76
+ },
77
+ "video_available": true,
78
+ "videos": [
79
+ {
80
+ "bytes": 4578041,
81
+ "channel": "spectator",
82
+ "fps": 10.0,
83
+ "frames": 543,
84
+ "path": "recording/third_person.mp4"
85
+ },
86
+ {
87
+ "bytes": 5576919,
88
+ "channel": "spectator",
89
+ "fps": 10.0,
90
+ "frames": 543,
91
+ "path": "recording/left_wrist.mp4"
92
+ },
93
+ {
94
+ "bytes": 2446160,
95
+ "channel": "spectator",
96
+ "fps": 10.0,
97
+ "frames": 543,
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-3/combined.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f30b26f5d8faf2959a29a4e92fa73238505196687f05eccbfe6dc543b032962
3
+ size 8255717
experiments/robotwin-native/kinex-3/detail.json ADDED
@@ -0,0 +1,1305 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "instruction": "use arms to grab the bottles and put them into the dustbin to the left of the table",
3
+ "observation": "Kinex wrote a measured Cartesian manipulation tool, picked and deposited the bottles sequentially, and reached native success at control frame 1,354.",
4
+ "usage": {
5
+ "n_input_tokens": 1644067,
6
+ "n_output_tokens": 10285,
7
+ "n_cache_tokens": 1598336,
8
+ "cost_usd": null,
9
+ "n_fresh_input_tokens": 45731,
10
+ "n_cache_read_tokens": 1598336,
11
+ "n_cache_write_tokens": 0
12
+ },
13
+ "usage_details": {
14
+ "accounting": "reported-responses",
15
+ "audit_complete": true,
16
+ "audit_source": "Provider terminal response usage",
17
+ "cache_hit_rate": 0.9721842236356547,
18
+ "cache_read_input_tokens": 1598336,
19
+ "cache_reported_input_tokens": 1644067,
20
+ "cache_write_basis": "provider-reported",
21
+ "cache_write_input_tokens": 0,
22
+ "cache_write_reported_input_tokens": 1644067,
23
+ "cached_input_tokens": 1598336,
24
+ "cost_usd": null,
25
+ "fresh_input_tokens": 45731,
26
+ "harbor_mismatches": {},
27
+ "input_breakdown_complete": true,
28
+ "input_tokens": 1644067,
29
+ "known_cache_write_input_tokens": 0,
30
+ "known_cached_input_tokens": 1598336,
31
+ "known_fresh_input_tokens": 45731,
32
+ "known_input_tokens": 1644067,
33
+ "known_output_tokens": 10285,
34
+ "known_reasoning_output_tokens": 2439,
35
+ "output_tokens": 10285,
36
+ "parser": "harbor/0.22.0",
37
+ "reasoning_output_tokens": 2439,
38
+ "reasoning_reported_output_tokens": 10285,
39
+ "rejected_attempts": 0,
40
+ "reported_responses": {
41
+ "cache_reported_input_tokens": 49,
42
+ "cache_write_input_tokens": 49,
43
+ "cache_write_reported_input_tokens": 49,
44
+ "cached_input_tokens": 49,
45
+ "fresh_input_tokens": 49,
46
+ "input_tokens": 49,
47
+ "output_tokens": 49,
48
+ "reasoning_output_tokens": 49,
49
+ "reasoning_reported_output_tokens": 49
50
+ },
51
+ "request_attempts": 49,
52
+ "response_count": 49,
53
+ "response_ids_complete": true,
54
+ "schema": "rlebench/token-usage/1",
55
+ "source": "Harbor session metrics",
56
+ "total_tokens": 1654352,
57
+ "uncached_input_tokens": 45731,
58
+ "unresolved_attempts": [],
59
+ "usage_records": [
60
+ {
61
+ "cache_write_input_tokens": 0,
62
+ "cached_input_tokens": 0,
63
+ "input_tokens": 11304,
64
+ "output_tokens": 19,
65
+ "reasoning_output_tokens": 0
66
+ },
67
+ {
68
+ "cache_write_input_tokens": 0,
69
+ "cached_input_tokens": 11136,
70
+ "input_tokens": 11344,
71
+ "output_tokens": 29,
72
+ "reasoning_output_tokens": 0
73
+ },
74
+ {
75
+ "cache_write_input_tokens": 0,
76
+ "cached_input_tokens": 11264,
77
+ "input_tokens": 11477,
78
+ "output_tokens": 30,
79
+ "reasoning_output_tokens": 0
80
+ },
81
+ {
82
+ "cache_write_input_tokens": 0,
83
+ "cached_input_tokens": 11392,
84
+ "input_tokens": 14222,
85
+ "output_tokens": 164,
86
+ "reasoning_output_tokens": 0
87
+ },
88
+ {
89
+ "cache_write_input_tokens": 0,
90
+ "cached_input_tokens": 14208,
91
+ "input_tokens": 17281,
92
+ "output_tokens": 173,
93
+ "reasoning_output_tokens": 0
94
+ },
95
+ {
96
+ "cache_write_input_tokens": 0,
97
+ "cached_input_tokens": 17280,
98
+ "input_tokens": 20636,
99
+ "output_tokens": 54,
100
+ "reasoning_output_tokens": 0
101
+ },
102
+ {
103
+ "cache_write_input_tokens": 0,
104
+ "cached_input_tokens": 20608,
105
+ "input_tokens": 21876,
106
+ "output_tokens": 98,
107
+ "reasoning_output_tokens": 0
108
+ },
109
+ {
110
+ "cache_write_input_tokens": 0,
111
+ "cached_input_tokens": 21888,
112
+ "input_tokens": 22170,
113
+ "output_tokens": 39,
114
+ "reasoning_output_tokens": 0
115
+ },
116
+ {
117
+ "cache_write_input_tokens": 0,
118
+ "cached_input_tokens": 22016,
119
+ "input_tokens": 22678,
120
+ "output_tokens": 23,
121
+ "reasoning_output_tokens": 0
122
+ },
123
+ {
124
+ "cache_write_input_tokens": 0,
125
+ "cached_input_tokens": 22528,
126
+ "input_tokens": 22870,
127
+ "output_tokens": 152,
128
+ "reasoning_output_tokens": 25
129
+ },
130
+ {
131
+ "cache_write_input_tokens": 0,
132
+ "cached_input_tokens": 22912,
133
+ "input_tokens": 24389,
134
+ "output_tokens": 362,
135
+ "reasoning_output_tokens": 0
136
+ },
137
+ {
138
+ "cache_write_input_tokens": 0,
139
+ "cached_input_tokens": 24576,
140
+ "input_tokens": 25147,
141
+ "output_tokens": 39,
142
+ "reasoning_output_tokens": 0
143
+ },
144
+ {
145
+ "cache_write_input_tokens": 0,
146
+ "cached_input_tokens": 25088,
147
+ "input_tokens": 25418,
148
+ "output_tokens": 354,
149
+ "reasoning_output_tokens": 0
150
+ },
151
+ {
152
+ "cache_write_input_tokens": 0,
153
+ "cached_input_tokens": 25600,
154
+ "input_tokens": 27068,
155
+ "output_tokens": 321,
156
+ "reasoning_output_tokens": 247
157
+ },
158
+ {
159
+ "cache_write_input_tokens": 0,
160
+ "cached_input_tokens": 27264,
161
+ "input_tokens": 28568,
162
+ "output_tokens": 1068,
163
+ "reasoning_output_tokens": 62
164
+ },
165
+ {
166
+ "cache_write_input_tokens": 0,
167
+ "cached_input_tokens": 29440,
168
+ "input_tokens": 29657,
169
+ "output_tokens": 74,
170
+ "reasoning_output_tokens": 0
171
+ },
172
+ {
173
+ "cache_write_input_tokens": 0,
174
+ "cached_input_tokens": 29568,
175
+ "input_tokens": 30414,
176
+ "output_tokens": 280,
177
+ "reasoning_output_tokens": 43
178
+ },
179
+ {
180
+ "cache_write_input_tokens": 0,
181
+ "cached_input_tokens": 30592,
182
+ "input_tokens": 30842,
183
+ "output_tokens": 252,
184
+ "reasoning_output_tokens": 0
185
+ },
186
+ {
187
+ "cache_write_input_tokens": 0,
188
+ "cached_input_tokens": 30976,
189
+ "input_tokens": 31389,
190
+ "output_tokens": 393,
191
+ "reasoning_output_tokens": 115
192
+ },
193
+ {
194
+ "cache_write_input_tokens": 0,
195
+ "cached_input_tokens": 31616,
196
+ "input_tokens": 32053,
197
+ "output_tokens": 189,
198
+ "reasoning_output_tokens": 94
199
+ },
200
+ {
201
+ "cache_write_input_tokens": 0,
202
+ "cached_input_tokens": 31872,
203
+ "input_tokens": 32718,
204
+ "output_tokens": 51,
205
+ "reasoning_output_tokens": 11
206
+ },
207
+ {
208
+ "cache_write_input_tokens": 0,
209
+ "cached_input_tokens": 32640,
210
+ "input_tokens": 32791,
211
+ "output_tokens": 277,
212
+ "reasoning_output_tokens": 46
213
+ },
214
+ {
215
+ "cache_write_input_tokens": 0,
216
+ "cached_input_tokens": 32896,
217
+ "input_tokens": 33530,
218
+ "output_tokens": 328,
219
+ "reasoning_output_tokens": 126
220
+ },
221
+ {
222
+ "cache_write_input_tokens": 0,
223
+ "cached_input_tokens": 33664,
224
+ "input_tokens": 34081,
225
+ "output_tokens": 283,
226
+ "reasoning_output_tokens": 142
227
+ },
228
+ {
229
+ "cache_write_input_tokens": 0,
230
+ "cached_input_tokens": 34176,
231
+ "input_tokens": 34391,
232
+ "output_tokens": 99,
233
+ "reasoning_output_tokens": 0
234
+ },
235
+ {
236
+ "cache_write_input_tokens": 0,
237
+ "cached_input_tokens": 34304,
238
+ "input_tokens": 35033,
239
+ "output_tokens": 134,
240
+ "reasoning_output_tokens": 0
241
+ },
242
+ {
243
+ "cache_write_input_tokens": 0,
244
+ "cached_input_tokens": 35072,
245
+ "input_tokens": 35705,
246
+ "output_tokens": 52,
247
+ "reasoning_output_tokens": 0
248
+ },
249
+ {
250
+ "cache_write_input_tokens": 0,
251
+ "cached_input_tokens": 35584,
252
+ "input_tokens": 36128,
253
+ "output_tokens": 88,
254
+ "reasoning_output_tokens": 0
255
+ },
256
+ {
257
+ "cache_write_input_tokens": 0,
258
+ "cached_input_tokens": 36096,
259
+ "input_tokens": 36756,
260
+ "output_tokens": 445,
261
+ "reasoning_output_tokens": 353
262
+ },
263
+ {
264
+ "cache_write_input_tokens": 0,
265
+ "cached_input_tokens": 36992,
266
+ "input_tokens": 37742,
267
+ "output_tokens": 155,
268
+ "reasoning_output_tokens": 64
269
+ },
270
+ {
271
+ "cache_write_input_tokens": 0,
272
+ "cached_input_tokens": 37760,
273
+ "input_tokens": 38332,
274
+ "output_tokens": 52,
275
+ "reasoning_output_tokens": 0
276
+ },
277
+ {
278
+ "cache_write_input_tokens": 0,
279
+ "cached_input_tokens": 38272,
280
+ "input_tokens": 38748,
281
+ "output_tokens": 179,
282
+ "reasoning_output_tokens": 78
283
+ },
284
+ {
285
+ "cache_write_input_tokens": 0,
286
+ "cached_input_tokens": 38784,
287
+ "input_tokens": 39356,
288
+ "output_tokens": 795,
289
+ "reasoning_output_tokens": 487
290
+ },
291
+ {
292
+ "cache_write_input_tokens": 0,
293
+ "cached_input_tokens": 40064,
294
+ "input_tokens": 40560,
295
+ "output_tokens": 412,
296
+ "reasoning_output_tokens": 214
297
+ },
298
+ {
299
+ "cache_write_input_tokens": 0,
300
+ "cached_input_tokens": 40832,
301
+ "input_tokens": 41227,
302
+ "output_tokens": 230,
303
+ "reasoning_output_tokens": 30
304
+ },
305
+ {
306
+ "cache_write_input_tokens": 0,
307
+ "cached_input_tokens": 41344,
308
+ "input_tokens": 41999,
309
+ "output_tokens": 52,
310
+ "reasoning_output_tokens": 0
311
+ },
312
+ {
313
+ "cache_write_input_tokens": 0,
314
+ "cached_input_tokens": 41856,
315
+ "input_tokens": 42422,
316
+ "output_tokens": 88,
317
+ "reasoning_output_tokens": 0
318
+ },
319
+ {
320
+ "cache_write_input_tokens": 0,
321
+ "cached_input_tokens": 42368,
322
+ "input_tokens": 43056,
323
+ "output_tokens": 130,
324
+ "reasoning_output_tokens": 0
325
+ },
326
+ {
327
+ "cache_write_input_tokens": 0,
328
+ "cached_input_tokens": 43136,
329
+ "input_tokens": 43618,
330
+ "output_tokens": 51,
331
+ "reasoning_output_tokens": 0
332
+ },
333
+ {
334
+ "cache_write_input_tokens": 0,
335
+ "cached_input_tokens": 43520,
336
+ "input_tokens": 44040,
337
+ "output_tokens": 128,
338
+ "reasoning_output_tokens": 30
339
+ },
340
+ {
341
+ "cache_write_input_tokens": 0,
342
+ "cached_input_tokens": 44032,
343
+ "input_tokens": 44597,
344
+ "output_tokens": 169,
345
+ "reasoning_output_tokens": 0
346
+ },
347
+ {
348
+ "cache_write_input_tokens": 0,
349
+ "cached_input_tokens": 44672,
350
+ "input_tokens": 45306,
351
+ "output_tokens": 52,
352
+ "reasoning_output_tokens": 0
353
+ },
354
+ {
355
+ "cache_write_input_tokens": 0,
356
+ "cached_input_tokens": 45184,
357
+ "input_tokens": 45725,
358
+ "output_tokens": 87,
359
+ "reasoning_output_tokens": 0
360
+ },
361
+ {
362
+ "cache_write_input_tokens": 0,
363
+ "cached_input_tokens": 45696,
364
+ "input_tokens": 46351,
365
+ "output_tokens": 132,
366
+ "reasoning_output_tokens": 0
367
+ },
368
+ {
369
+ "cache_write_input_tokens": 0,
370
+ "cached_input_tokens": 46464,
371
+ "input_tokens": 46912,
372
+ "output_tokens": 53,
373
+ "reasoning_output_tokens": 0
374
+ },
375
+ {
376
+ "cache_write_input_tokens": 0,
377
+ "cached_input_tokens": 46848,
378
+ "input_tokens": 47103,
379
+ "output_tokens": 93,
380
+ "reasoning_output_tokens": 42
381
+ },
382
+ {
383
+ "cache_write_input_tokens": 0,
384
+ "cached_input_tokens": 47104,
385
+ "input_tokens": 47579,
386
+ "output_tokens": 695,
387
+ "reasoning_output_tokens": 230
388
+ },
389
+ {
390
+ "cache_write_input_tokens": 0,
391
+ "cached_input_tokens": 48128,
392
+ "input_tokens": 48301,
393
+ "output_tokens": 829,
394
+ "reasoning_output_tokens": 0
395
+ },
396
+ {
397
+ "cache_write_input_tokens": 0,
398
+ "cached_input_tokens": 49024,
399
+ "input_tokens": 49157,
400
+ "output_tokens": 33,
401
+ "reasoning_output_tokens": 0
402
+ }
403
+ ],
404
+ "usage_source": "provider-raw"
405
+ },
406
+ "cost_note": "API equivalent sums the native Pi session cost estimates. Subscription billing is not reported.",
407
+ "media_note": "All three native views and every frame retained. Frame timing is unchanged; playback starts at zero.",
408
+ "artifacts": [
409
+ {
410
+ "name": "artifacts.json",
411
+ "path": "experiments/robotwin-native/kinex-3/artifacts.json",
412
+ "bytes": 2655,
413
+ "sha256": "2e3b7f8171c0aa4d53500c19d10624f0423eb7fbeef544c8f74fc125fbe13453"
414
+ },
415
+ {
416
+ "name": "combined.mp4",
417
+ "path": "experiments/robotwin-native/kinex-3/combined.mp4",
418
+ "bytes": 8255717,
419
+ "sha256": "8f30b26f5d8faf2959a29a4e92fa73238505196687f05eccbfe6dc543b032962"
420
+ },
421
+ {
422
+ "name": "effective-model.json",
423
+ "path": "experiments/robotwin-native/kinex-3/effective-model.json",
424
+ "bytes": 444,
425
+ "sha256": "5a3694bfdd4fbeb4b1f47e125dde01703a7e004ec43f0bd6b47bbe7ba7c989e5"
426
+ },
427
+ {
428
+ "name": "episode-report.json",
429
+ "path": "experiments/robotwin-native/kinex-3/episode-report.json",
430
+ "bytes": 12888,
431
+ "sha256": "c0134794182e5cbc5dee04e2f02ed347c13496d9187172f784bbdd389a54a238"
432
+ },
433
+ {
434
+ "name": "episode.json",
435
+ "path": "experiments/robotwin-native/kinex-3/episode.json",
436
+ "bytes": 293,
437
+ "sha256": "7c00c5ac10e73269fdffa5b67e61df1f64e0bd95597e8264eabf8d7c73cc7c6a"
438
+ },
439
+ {
440
+ "name": "evidence.tar.gz",
441
+ "path": "experiments/robotwin-native/kinex-3/evidence.tar.gz",
442
+ "bytes": 12619391,
443
+ "sha256": "2d3884a8c19a136e796d8a761058c2dcb24b467aef261ab6e3872c8b91f4c487"
444
+ },
445
+ {
446
+ "name": "instruction.json",
447
+ "path": "experiments/robotwin-native/kinex-3/instruction.json",
448
+ "bytes": 4903,
449
+ "sha256": "39fdfb718fe2f7d1a0f3b9fbc674ac5207bc057d8ab7ac621d8ea561f9353e38"
450
+ },
451
+ {
452
+ "name": "media.json",
453
+ "path": "experiments/robotwin-native/kinex-3/media.json",
454
+ "bytes": 2325,
455
+ "sha256": "099df283ee808473ee8af1f89d6da2de8c25cb0d875bd892d4d24bbae39d7d71"
456
+ },
457
+ {
458
+ "name": "observations/1c00b1c6-1.png",
459
+ "path": "experiments/robotwin-native/kinex-3/observations/1c00b1c6-1.png",
460
+ "bytes": 60737,
461
+ "sha256": "7063b0ccdb1dbc2d02e29e82543cce6a5d5c327804cbbc777580895bc5d98dbd"
462
+ },
463
+ {
464
+ "name": "observations/1c00b1c6-2.png",
465
+ "path": "experiments/robotwin-native/kinex-3/observations/1c00b1c6-2.png",
466
+ "bytes": 61831,
467
+ "sha256": "fbb5da663e7d017be22f35d45383db8348fbf59780c18b7ba142aae4d558266a"
468
+ },
469
+ {
470
+ "name": "observations/25fb6783-1.png",
471
+ "path": "experiments/robotwin-native/kinex-3/observations/25fb6783-1.png",
472
+ "bytes": 68415,
473
+ "sha256": "e3d1c3e98145951111368baf940997e1ea724580af7c623abacceebb050cd59c"
474
+ },
475
+ {
476
+ "name": "observations/2e34d545-1.png",
477
+ "path": "experiments/robotwin-native/kinex-3/observations/2e34d545-1.png",
478
+ "bytes": 40375,
479
+ "sha256": "7f34487125c37f41b9a29898f1ec1d997abc8bd87e0487b88f58cc0f9b161dba"
480
+ },
481
+ {
482
+ "name": "observations/38f9e622-1.png",
483
+ "path": "experiments/robotwin-native/kinex-3/observations/38f9e622-1.png",
484
+ "bytes": 65294,
485
+ "sha256": "03e6d0502942bff9ed406b8f2b8094e814d2792c674baa0e4d3fded0078a910e"
486
+ },
487
+ {
488
+ "name": "observations/38f9e622-2.png",
489
+ "path": "experiments/robotwin-native/kinex-3/observations/38f9e622-2.png",
490
+ "bytes": 61357,
491
+ "sha256": "1b9c255734eeb6ce429a8827f8a3ceb611f15b4dbe386a07462f19ecca9a6797"
492
+ },
493
+ {
494
+ "name": "observations/40cd7359-1.png",
495
+ "path": "experiments/robotwin-native/kinex-3/observations/40cd7359-1.png",
496
+ "bytes": 36978,
497
+ "sha256": "44a7d05048edafee2ee2157849a02b6f392106fbdaf9abd57c79276727ec7306"
498
+ },
499
+ {
500
+ "name": "observations/4a031853-1.png",
501
+ "path": "experiments/robotwin-native/kinex-3/observations/4a031853-1.png",
502
+ "bytes": 68025,
503
+ "sha256": "cac6672ac13d068e53202cb7deb504d316a5df5336aa3ba516e15a399361fbbc"
504
+ },
505
+ {
506
+ "name": "observations/4a031853-2.png",
507
+ "path": "experiments/robotwin-native/kinex-3/observations/4a031853-2.png",
508
+ "bytes": 58041,
509
+ "sha256": "794cd80fb7052561ebf1bf3f7eeeaecde86ebc1ba6d593e021d02d501fb4ce8f"
510
+ },
511
+ {
512
+ "name": "observations/5306339c-1.png",
513
+ "path": "experiments/robotwin-native/kinex-3/observations/5306339c-1.png",
514
+ "bytes": 39587,
515
+ "sha256": "b0f3557c417f70c4d62f197fbb108c3ae14a44ae655b8999b03ce335b0ad5ade"
516
+ },
517
+ {
518
+ "name": "observations/5306339c-2.png",
519
+ "path": "experiments/robotwin-native/kinex-3/observations/5306339c-2.png",
520
+ "bytes": 39496,
521
+ "sha256": "cb9056c0b4026360e7a7a318c758248eb162018ed05510269e2b9d4192baf61a"
522
+ },
523
+ {
524
+ "name": "observations/59f4e1b6-1.png",
525
+ "path": "experiments/robotwin-native/kinex-3/observations/59f4e1b6-1.png",
526
+ "bytes": 76369,
527
+ "sha256": "6f2aff26367dba34e54efa6ec7debc7fd9dbca1a61f51b71d23e6013d13a62cd"
528
+ },
529
+ {
530
+ "name": "observations/7af7cb21-1.png",
531
+ "path": "experiments/robotwin-native/kinex-3/observations/7af7cb21-1.png",
532
+ "bytes": 76618,
533
+ "sha256": "0fbbe4bdc88b3e8920da022d3432d091d0f7a9bb401d73da333d60433fed9011"
534
+ },
535
+ {
536
+ "name": "observations/82847509-1.png",
537
+ "path": "experiments/robotwin-native/kinex-3/observations/82847509-1.png",
538
+ "bytes": 77003,
539
+ "sha256": "b449f3aae9e2737c3651dec459330f6ed359c82f7fe64c908443ae232fd679f5"
540
+ },
541
+ {
542
+ "name": "observations/8c8cb210-1.png",
543
+ "path": "experiments/robotwin-native/kinex-3/observations/8c8cb210-1.png",
544
+ "bytes": 66630,
545
+ "sha256": "bcecd45c2146cce6e26ec6f1df4be1c7b24d63b03e128dec5f07ff2fe9f68d35"
546
+ },
547
+ {
548
+ "name": "observations/8c8cb210-2.png",
549
+ "path": "experiments/robotwin-native/kinex-3/observations/8c8cb210-2.png",
550
+ "bytes": 34234,
551
+ "sha256": "ca9d5e37554f7530a0279c39ffc40c5409a8d0b3f129d8a4bc37f0c611ff1393"
552
+ },
553
+ {
554
+ "name": "observations/9deb50d9-1.png",
555
+ "path": "experiments/robotwin-native/kinex-3/observations/9deb50d9-1.png",
556
+ "bytes": 77695,
557
+ "sha256": "2f7c37774aa11ff05c26a2a2fa70ede6b602afb1364c213056ed273bb97b53b7"
558
+ },
559
+ {
560
+ "name": "observations/9deb50d9-2.png",
561
+ "path": "experiments/robotwin-native/kinex-3/observations/9deb50d9-2.png",
562
+ "bytes": 45045,
563
+ "sha256": "71a92dd42f794aa63513a285ef0941cd5ca4b891a8ba72716a591785f4b8c4db"
564
+ },
565
+ {
566
+ "name": "observations/a2501cc5-1.png",
567
+ "path": "experiments/robotwin-native/kinex-3/observations/a2501cc5-1.png",
568
+ "bytes": 79335,
569
+ "sha256": "7c5aeb0bc313a1a6eadc82017525c63b3ddef87e256434f3e462b6fd11acef80"
570
+ },
571
+ {
572
+ "name": "observations/a57d6a5b-1.png",
573
+ "path": "experiments/robotwin-native/kinex-3/observations/a57d6a5b-1.png",
574
+ "bytes": 85004,
575
+ "sha256": "b08e50adfecc62b41d3d583e49928efe02094602f1cb4198e793e03e71d9653b"
576
+ },
577
+ {
578
+ "name": "observations/b27b5e13-1.png",
579
+ "path": "experiments/robotwin-native/kinex-3/observations/b27b5e13-1.png",
580
+ "bytes": 56960,
581
+ "sha256": "6b729687f861f086aa5e2f2c967afb06420e8911e80233bfb3b69e2699a1bee7"
582
+ },
583
+ {
584
+ "name": "observations/bd40c7f2-1.png",
585
+ "path": "experiments/robotwin-native/kinex-3/observations/bd40c7f2-1.png",
586
+ "bytes": 66127,
587
+ "sha256": "3afe3997a3bf2d6a66907aac41c76caf353c2bc068f1120826d7c7e5ba8c07a5"
588
+ },
589
+ {
590
+ "name": "observations/bd40c7f2-2.png",
591
+ "path": "experiments/robotwin-native/kinex-3/observations/bd40c7f2-2.png",
592
+ "bytes": 42241,
593
+ "sha256": "a4befa13ce8d924eac356818f9b80b96a775bcac0357aaaa7383d3f169d38403"
594
+ },
595
+ {
596
+ "name": "observations/e0aba462-1.png",
597
+ "path": "experiments/robotwin-native/kinex-3/observations/e0aba462-1.png",
598
+ "bytes": 50874,
599
+ "sha256": "055357a284b4e5eec8df851563e9dda8fae19e1319e29ce3b4e706a56be05db2"
600
+ },
601
+ {
602
+ "name": "observations/e8dcb044-1.png",
603
+ "path": "experiments/robotwin-native/kinex-3/observations/e8dcb044-1.png",
604
+ "bytes": 76964,
605
+ "sha256": "ab5d971fa2eb24d16c49a20b4df0a7300a4cf726a7e218805f55740e3d56a046"
606
+ },
607
+ {
608
+ "name": "observations/e8dcb044-2.png",
609
+ "path": "experiments/robotwin-native/kinex-3/observations/e8dcb044-2.png",
610
+ "bytes": 69817,
611
+ "sha256": "0cb4fa8d3a474c64226501fb82102facc644d3216fbb6d055ecc3b6922b48587"
612
+ },
613
+ {
614
+ "name": "observations/ed3b0775-1.png",
615
+ "path": "experiments/robotwin-native/kinex-3/observations/ed3b0775-1.png",
616
+ "bytes": 77329,
617
+ "sha256": "52f30be92a7ed506ce39b7188ef68bee65fa3cf1b4461dc865893a414c2e1dc8"
618
+ },
619
+ {
620
+ "name": "observations/ed3b0775-2.png",
621
+ "path": "experiments/robotwin-native/kinex-3/observations/ed3b0775-2.png",
622
+ "bytes": 62651,
623
+ "sha256": "b9ed6c867d46b3d4074b534432f62e579ea56540df8a55564ce8313d53e44cab"
624
+ },
625
+ {
626
+ "name": "observations/eda38a00-1.png",
627
+ "path": "experiments/robotwin-native/kinex-3/observations/eda38a00-1.png",
628
+ "bytes": 66202,
629
+ "sha256": "bae20dd5493df8040b2105584a6dca5eb94530de18ab9616b6b8f860c04ab990"
630
+ },
631
+ {
632
+ "name": "observations/eda38a00-2.png",
633
+ "path": "experiments/robotwin-native/kinex-3/observations/eda38a00-2.png",
634
+ "bytes": 74831,
635
+ "sha256": "cd84f4c4014afd27254d713b6b8a8506ee90b85f356a4b9aec52af91b3ec1174"
636
+ },
637
+ {
638
+ "name": "observations/ff2d32db-1.png",
639
+ "path": "experiments/robotwin-native/kinex-3/observations/ff2d32db-1.png",
640
+ "bytes": 37211,
641
+ "sha256": "b6ce0e0ebfe1b31fce9d088565c0636cb86b64e428bbeb150905ba9bd26dd64b"
642
+ },
643
+ {
644
+ "name": "policy-calls.json",
645
+ "path": "experiments/robotwin-native/kinex-3/policy-calls.json",
646
+ "bytes": 3,
647
+ "sha256": "37517e5f3dc66819f61f5a7bb8ace1921282415f10551d2defa5c3eb0985b570"
648
+ },
649
+ {
650
+ "name": "poster.jpg",
651
+ "path": "experiments/robotwin-native/kinex-3/poster.jpg",
652
+ "bytes": 40772,
653
+ "sha256": "7cfa89f6915dd66f0aead3d4ef82e1a961f6ad52eca698ee744e3b2c88e618a9"
654
+ },
655
+ {
656
+ "name": "provider-requests.jsonl",
657
+ "path": "experiments/robotwin-native/kinex-3/provider-requests.jsonl",
658
+ "bytes": 937385,
659
+ "sha256": "d43016f699570420014b74a7679e9ecbd1ce66d9d276a8bb6fe047eab2e007e7"
660
+ },
661
+ {
662
+ "name": "session.json",
663
+ "path": "experiments/robotwin-native/kinex-3/session.json",
664
+ "bytes": 128265,
665
+ "sha256": "821b8a18b9e418284bb68cd70a3b76cd62767b9a9bc59fa5429de07067e1346b"
666
+ },
667
+ {
668
+ "name": "session.jsonl",
669
+ "path": "experiments/robotwin-native/kinex-3/session.jsonl",
670
+ "bytes": 2837484,
671
+ "sha256": "89d426f9f6eb92b70abda6c1478417c64dd19aa172af01ecadacade2b8dbfbb1"
672
+ },
673
+ {
674
+ "name": "terminal.jpg",
675
+ "path": "experiments/robotwin-native/kinex-3/terminal.jpg",
676
+ "bytes": 27941,
677
+ "sha256": "f25695a48e74ad45cd46f34ae0907336b2b6da703036cab3a35ad14bfe7c5785"
678
+ },
679
+ {
680
+ "name": "tool-calls.json",
681
+ "path": "experiments/robotwin-native/kinex-3/tool-calls.json",
682
+ "bytes": 124797,
683
+ "sha256": "403b525738a71f3452ad1c06a81b377f8d8e2fc61d43c91a37e8958763de169a"
684
+ },
685
+ {
686
+ "name": "trace.json",
687
+ "path": "experiments/robotwin-native/kinex-3/trace.json",
688
+ "bytes": 237758,
689
+ "sha256": "6a13c27c1ff901be5ef7c5d30d8a1c13883ad01b67232561a08875bcd9af07df"
690
+ },
691
+ {
692
+ "name": "usage-audit.json",
693
+ "path": "experiments/robotwin-native/kinex-3/usage-audit.json",
694
+ "bytes": 10295,
695
+ "sha256": "2614039697c7fd8ff404397847853f9b4d7c4b37254a951e3e53e7bf82d3ddef"
696
+ },
697
+ {
698
+ "name": "workspace/.kinex/devices/head_camera/profile.yaml",
699
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/head_camera/profile.yaml",
700
+ "bytes": 2691,
701
+ "sha256": "8554082b045e1b66ed92947f996bd323c2bd5b1bec6cf494b92597bfdc5d1aed"
702
+ },
703
+ {
704
+ "name": "workspace/.kinex/devices/left_arm/arm.urdf",
705
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_arm/arm.urdf",
706
+ "bytes": 1879,
707
+ "sha256": "e53c41aaf34a08cb0224e5bf2bf9bbf949813018c2e0091b8021ef04d82a3f41"
708
+ },
709
+ {
710
+ "name": "workspace/.kinex/devices/left_arm/links.py",
711
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_arm/links.py",
712
+ "bytes": 233,
713
+ "sha256": "6858db0b49065b895a79860ada359fd6da79a32e4c5b2950d159678420e85844"
714
+ },
715
+ {
716
+ "name": "workspace/.kinex/devices/left_arm/profile.yaml",
717
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_arm/profile.yaml",
718
+ "bytes": 2714,
719
+ "sha256": "54b56b1d4c8b7493c38e25cd42dbe530a17d758c8bb5b8829c1d88597afc5d29"
720
+ },
721
+ {
722
+ "name": "workspace/.kinex/devices/left_camera/profile.yaml",
723
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_camera/profile.yaml",
724
+ "bytes": 2734,
725
+ "sha256": "5d816e599d5b4e2fb99f4a8c18ba05b79784f08edb8ca20d2d22df806512973b"
726
+ },
727
+ {
728
+ "name": "workspace/.kinex/devices/left_gripper/gripper.urdf",
729
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_gripper/gripper.urdf",
730
+ "bytes": 1401,
731
+ "sha256": "7f5aad8ae559c5196bb28a9d72d7020bf5d7b79d5af49aefdb0e6ce279e1f4be"
732
+ },
733
+ {
734
+ "name": "workspace/.kinex/devices/left_gripper/links.py",
735
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_gripper/links.py",
736
+ "bytes": 233,
737
+ "sha256": "6858db0b49065b895a79860ada359fd6da79a32e4c5b2950d159678420e85844"
738
+ },
739
+ {
740
+ "name": "workspace/.kinex/devices/left_gripper/profile.yaml",
741
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/left_gripper/profile.yaml",
742
+ "bytes": 2594,
743
+ "sha256": "ae580279c93df2ebd7c5d038dcf7a9cc19a7541999c56643c2a5831411033f54"
744
+ },
745
+ {
746
+ "name": "workspace/.kinex/devices/right_arm/arm.urdf",
747
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_arm/arm.urdf",
748
+ "bytes": 1881,
749
+ "sha256": "cb210121ba099341eeffdcc95588deaa0524037cff46602940d1037628600d8e"
750
+ },
751
+ {
752
+ "name": "workspace/.kinex/devices/right_arm/links.py",
753
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_arm/links.py",
754
+ "bytes": 233,
755
+ "sha256": "6858db0b49065b895a79860ada359fd6da79a32e4c5b2950d159678420e85844"
756
+ },
757
+ {
758
+ "name": "workspace/.kinex/devices/right_arm/profile.yaml",
759
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_arm/profile.yaml",
760
+ "bytes": 2728,
761
+ "sha256": "309509cbbb1e70a3f6540a793cdac0f0926f723a8fd603cdad3c0a3b0dba2ab7"
762
+ },
763
+ {
764
+ "name": "workspace/.kinex/devices/right_camera/profile.yaml",
765
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_camera/profile.yaml",
766
+ "bytes": 2742,
767
+ "sha256": "75af946baa024823eed57ce75c0b5edb8f4611e3f446476921419ecd4f71cd96"
768
+ },
769
+ {
770
+ "name": "workspace/.kinex/devices/right_gripper/gripper.urdf",
771
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_gripper/gripper.urdf",
772
+ "bytes": 1407,
773
+ "sha256": "347f6969d8189abeee10b5a3e5358b91511858da9fb233cdf062f80c53b97dd9"
774
+ },
775
+ {
776
+ "name": "workspace/.kinex/devices/right_gripper/links.py",
777
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_gripper/links.py",
778
+ "bytes": 233,
779
+ "sha256": "6858db0b49065b895a79860ada359fd6da79a32e4c5b2950d159678420e85844"
780
+ },
781
+ {
782
+ "name": "workspace/.kinex/devices/right_gripper/profile.yaml",
783
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/right_gripper/profile.yaml",
784
+ "bytes": 2614,
785
+ "sha256": "a8725a54dc2e431ae9c14c18c383051d58148b466de6eb731d585beee0ad8975"
786
+ },
787
+ {
788
+ "name": "workspace/.kinex/devices/world/profile.yaml",
789
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/devices/world/profile.yaml",
790
+ "bytes": 1593,
791
+ "sha256": "cee6b114af8e433142fc7af636bb47bb8daab30356684974b0347b2e61d1dbbc"
792
+ },
793
+ {
794
+ "name": "workspace/.kinex/kinex.json",
795
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/kinex.json",
796
+ "bytes": 50,
797
+ "sha256": "703ec9cee9365c7c5c4b967219de9a9e7b9ca17517eae82cf7af30431de4026e"
798
+ },
799
+ {
800
+ "name": "workspace/.kinex/memos/put-bottles-dustbin.md",
801
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/memos/put-bottles-dustbin.md",
802
+ "bytes": 2943,
803
+ "sha256": "eddaa934c6d58f5d0504a470b3b424782b29255a72e79dfdd1e9386db19e53f4"
804
+ },
805
+ {
806
+ "name": "workspace/.kinex/tools/manipulate.py",
807
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/tools/manipulate.py",
808
+ "bytes": 4135,
809
+ "sha256": "04c3805880622dcf00c5d36453e8b31b5f831ea61708b618146b851ced21a804"
810
+ },
811
+ {
812
+ "name": "workspace/.kinex/workspace.yaml",
813
+ "path": "experiments/robotwin-native/kinex-3/workspace/.kinex/workspace.yaml",
814
+ "bytes": 335,
815
+ "sha256": "8a0cbe7e2438339e12693ecdda4530b8601e092a11c9e1d5f6301bd67a415ca7"
816
+ },
817
+ {
818
+ "name": "workspace/harness.md",
819
+ "path": "experiments/robotwin-native/kinex-3/workspace/harness.md",
820
+ "bytes": 444,
821
+ "sha256": "81c1ef513247775681eb49b5ac07a5f4eebc7848afab963ef21a3bca473b8485"
822
+ },
823
+ {
824
+ "name": "workspace/instruction.md",
825
+ "path": "experiments/robotwin-native/kinex-3/workspace/instruction.md",
826
+ "bytes": 2157,
827
+ "sha256": "193190387438d3d7ba22ab897536f92e193bef98c94ec09769e371e18b54b1f4"
828
+ },
829
+ {
830
+ "name": "workspace.json",
831
+ "path": "experiments/robotwin-native/kinex-3/workspace.json",
832
+ "bytes": 4630,
833
+ "sha256": "1509614267f04fc29b8cb269f1d91da0acf1ce4c775377b11f5e858be23a0bf2"
834
+ },
835
+ {
836
+ "name": "workspace.tar.gz",
837
+ "path": "experiments/robotwin-native/kinex-3/workspace.tar.gz",
838
+ "bytes": 9821,
839
+ "sha256": "bb3782a283913dd8ece0ae26fd20814e479c6c2a7c9d76c335f3b8339f3e113a"
840
+ }
841
+ ],
842
+ "native": {
843
+ "episode": 1,
844
+ "initialization": {
845
+ "expert_plan_success": true,
846
+ "requested_seed": 0
847
+ },
848
+ "native_reward": 1.0,
849
+ "planned_episodes": 1,
850
+ "reason": "success",
851
+ "simulation_time_s": 54.160000000000004,
852
+ "step_limit": 1700,
853
+ "steps": 1354,
854
+ "success": true,
855
+ "valid": true,
856
+ "step": "episode_01",
857
+ "evidence_valid": true,
858
+ "wall_time_sec": 756.194745,
859
+ "usage": {
860
+ "n_input_tokens": 1644067,
861
+ "n_output_tokens": 10285,
862
+ "n_cache_tokens": 1598336,
863
+ "cost_usd": null,
864
+ "n_fresh_input_tokens": 45731,
865
+ "n_cache_read_tokens": 1598336,
866
+ "n_cache_write_tokens": 0
867
+ },
868
+ "usage_details": {
869
+ "accounting": "reported-responses",
870
+ "audit_complete": true,
871
+ "audit_source": "Provider terminal response usage",
872
+ "cache_hit_rate": 0.9721842236356547,
873
+ "cache_read_input_tokens": 1598336,
874
+ "cache_reported_input_tokens": 1644067,
875
+ "cache_write_basis": "provider-reported",
876
+ "cache_write_input_tokens": 0,
877
+ "cache_write_reported_input_tokens": 1644067,
878
+ "cached_input_tokens": 1598336,
879
+ "cost_usd": null,
880
+ "fresh_input_tokens": 45731,
881
+ "harbor_mismatches": {},
882
+ "input_breakdown_complete": true,
883
+ "input_tokens": 1644067,
884
+ "known_cache_write_input_tokens": 0,
885
+ "known_cached_input_tokens": 1598336,
886
+ "known_fresh_input_tokens": 45731,
887
+ "known_input_tokens": 1644067,
888
+ "known_output_tokens": 10285,
889
+ "known_reasoning_output_tokens": 2439,
890
+ "output_tokens": 10285,
891
+ "parser": "harbor/0.22.0",
892
+ "reasoning_output_tokens": 2439,
893
+ "reasoning_reported_output_tokens": 10285,
894
+ "rejected_attempts": 0,
895
+ "reported_responses": {
896
+ "cache_reported_input_tokens": 49,
897
+ "cache_write_input_tokens": 49,
898
+ "cache_write_reported_input_tokens": 49,
899
+ "cached_input_tokens": 49,
900
+ "fresh_input_tokens": 49,
901
+ "input_tokens": 49,
902
+ "output_tokens": 49,
903
+ "reasoning_output_tokens": 49,
904
+ "reasoning_reported_output_tokens": 49
905
+ },
906
+ "request_attempts": 49,
907
+ "response_count": 49,
908
+ "response_ids_complete": true,
909
+ "schema": "rlebench/token-usage/1",
910
+ "source": "Harbor session metrics",
911
+ "total_tokens": 1654352,
912
+ "uncached_input_tokens": 45731,
913
+ "unresolved_attempts": [],
914
+ "usage_records": [
915
+ {
916
+ "cache_write_input_tokens": 0,
917
+ "cached_input_tokens": 0,
918
+ "input_tokens": 11304,
919
+ "output_tokens": 19,
920
+ "reasoning_output_tokens": 0
921
+ },
922
+ {
923
+ "cache_write_input_tokens": 0,
924
+ "cached_input_tokens": 11136,
925
+ "input_tokens": 11344,
926
+ "output_tokens": 29,
927
+ "reasoning_output_tokens": 0
928
+ },
929
+ {
930
+ "cache_write_input_tokens": 0,
931
+ "cached_input_tokens": 11264,
932
+ "input_tokens": 11477,
933
+ "output_tokens": 30,
934
+ "reasoning_output_tokens": 0
935
+ },
936
+ {
937
+ "cache_write_input_tokens": 0,
938
+ "cached_input_tokens": 11392,
939
+ "input_tokens": 14222,
940
+ "output_tokens": 164,
941
+ "reasoning_output_tokens": 0
942
+ },
943
+ {
944
+ "cache_write_input_tokens": 0,
945
+ "cached_input_tokens": 14208,
946
+ "input_tokens": 17281,
947
+ "output_tokens": 173,
948
+ "reasoning_output_tokens": 0
949
+ },
950
+ {
951
+ "cache_write_input_tokens": 0,
952
+ "cached_input_tokens": 17280,
953
+ "input_tokens": 20636,
954
+ "output_tokens": 54,
955
+ "reasoning_output_tokens": 0
956
+ },
957
+ {
958
+ "cache_write_input_tokens": 0,
959
+ "cached_input_tokens": 20608,
960
+ "input_tokens": 21876,
961
+ "output_tokens": 98,
962
+ "reasoning_output_tokens": 0
963
+ },
964
+ {
965
+ "cache_write_input_tokens": 0,
966
+ "cached_input_tokens": 21888,
967
+ "input_tokens": 22170,
968
+ "output_tokens": 39,
969
+ "reasoning_output_tokens": 0
970
+ },
971
+ {
972
+ "cache_write_input_tokens": 0,
973
+ "cached_input_tokens": 22016,
974
+ "input_tokens": 22678,
975
+ "output_tokens": 23,
976
+ "reasoning_output_tokens": 0
977
+ },
978
+ {
979
+ "cache_write_input_tokens": 0,
980
+ "cached_input_tokens": 22528,
981
+ "input_tokens": 22870,
982
+ "output_tokens": 152,
983
+ "reasoning_output_tokens": 25
984
+ },
985
+ {
986
+ "cache_write_input_tokens": 0,
987
+ "cached_input_tokens": 22912,
988
+ "input_tokens": 24389,
989
+ "output_tokens": 362,
990
+ "reasoning_output_tokens": 0
991
+ },
992
+ {
993
+ "cache_write_input_tokens": 0,
994
+ "cached_input_tokens": 24576,
995
+ "input_tokens": 25147,
996
+ "output_tokens": 39,
997
+ "reasoning_output_tokens": 0
998
+ },
999
+ {
1000
+ "cache_write_input_tokens": 0,
1001
+ "cached_input_tokens": 25088,
1002
+ "input_tokens": 25418,
1003
+ "output_tokens": 354,
1004
+ "reasoning_output_tokens": 0
1005
+ },
1006
+ {
1007
+ "cache_write_input_tokens": 0,
1008
+ "cached_input_tokens": 25600,
1009
+ "input_tokens": 27068,
1010
+ "output_tokens": 321,
1011
+ "reasoning_output_tokens": 247
1012
+ },
1013
+ {
1014
+ "cache_write_input_tokens": 0,
1015
+ "cached_input_tokens": 27264,
1016
+ "input_tokens": 28568,
1017
+ "output_tokens": 1068,
1018
+ "reasoning_output_tokens": 62
1019
+ },
1020
+ {
1021
+ "cache_write_input_tokens": 0,
1022
+ "cached_input_tokens": 29440,
1023
+ "input_tokens": 29657,
1024
+ "output_tokens": 74,
1025
+ "reasoning_output_tokens": 0
1026
+ },
1027
+ {
1028
+ "cache_write_input_tokens": 0,
1029
+ "cached_input_tokens": 29568,
1030
+ "input_tokens": 30414,
1031
+ "output_tokens": 280,
1032
+ "reasoning_output_tokens": 43
1033
+ },
1034
+ {
1035
+ "cache_write_input_tokens": 0,
1036
+ "cached_input_tokens": 30592,
1037
+ "input_tokens": 30842,
1038
+ "output_tokens": 252,
1039
+ "reasoning_output_tokens": 0
1040
+ },
1041
+ {
1042
+ "cache_write_input_tokens": 0,
1043
+ "cached_input_tokens": 30976,
1044
+ "input_tokens": 31389,
1045
+ "output_tokens": 393,
1046
+ "reasoning_output_tokens": 115
1047
+ },
1048
+ {
1049
+ "cache_write_input_tokens": 0,
1050
+ "cached_input_tokens": 31616,
1051
+ "input_tokens": 32053,
1052
+ "output_tokens": 189,
1053
+ "reasoning_output_tokens": 94
1054
+ },
1055
+ {
1056
+ "cache_write_input_tokens": 0,
1057
+ "cached_input_tokens": 31872,
1058
+ "input_tokens": 32718,
1059
+ "output_tokens": 51,
1060
+ "reasoning_output_tokens": 11
1061
+ },
1062
+ {
1063
+ "cache_write_input_tokens": 0,
1064
+ "cached_input_tokens": 32640,
1065
+ "input_tokens": 32791,
1066
+ "output_tokens": 277,
1067
+ "reasoning_output_tokens": 46
1068
+ },
1069
+ {
1070
+ "cache_write_input_tokens": 0,
1071
+ "cached_input_tokens": 32896,
1072
+ "input_tokens": 33530,
1073
+ "output_tokens": 328,
1074
+ "reasoning_output_tokens": 126
1075
+ },
1076
+ {
1077
+ "cache_write_input_tokens": 0,
1078
+ "cached_input_tokens": 33664,
1079
+ "input_tokens": 34081,
1080
+ "output_tokens": 283,
1081
+ "reasoning_output_tokens": 142
1082
+ },
1083
+ {
1084
+ "cache_write_input_tokens": 0,
1085
+ "cached_input_tokens": 34176,
1086
+ "input_tokens": 34391,
1087
+ "output_tokens": 99,
1088
+ "reasoning_output_tokens": 0
1089
+ },
1090
+ {
1091
+ "cache_write_input_tokens": 0,
1092
+ "cached_input_tokens": 34304,
1093
+ "input_tokens": 35033,
1094
+ "output_tokens": 134,
1095
+ "reasoning_output_tokens": 0
1096
+ },
1097
+ {
1098
+ "cache_write_input_tokens": 0,
1099
+ "cached_input_tokens": 35072,
1100
+ "input_tokens": 35705,
1101
+ "output_tokens": 52,
1102
+ "reasoning_output_tokens": 0
1103
+ },
1104
+ {
1105
+ "cache_write_input_tokens": 0,
1106
+ "cached_input_tokens": 35584,
1107
+ "input_tokens": 36128,
1108
+ "output_tokens": 88,
1109
+ "reasoning_output_tokens": 0
1110
+ },
1111
+ {
1112
+ "cache_write_input_tokens": 0,
1113
+ "cached_input_tokens": 36096,
1114
+ "input_tokens": 36756,
1115
+ "output_tokens": 445,
1116
+ "reasoning_output_tokens": 353
1117
+ },
1118
+ {
1119
+ "cache_write_input_tokens": 0,
1120
+ "cached_input_tokens": 36992,
1121
+ "input_tokens": 37742,
1122
+ "output_tokens": 155,
1123
+ "reasoning_output_tokens": 64
1124
+ },
1125
+ {
1126
+ "cache_write_input_tokens": 0,
1127
+ "cached_input_tokens": 37760,
1128
+ "input_tokens": 38332,
1129
+ "output_tokens": 52,
1130
+ "reasoning_output_tokens": 0
1131
+ },
1132
+ {
1133
+ "cache_write_input_tokens": 0,
1134
+ "cached_input_tokens": 38272,
1135
+ "input_tokens": 38748,
1136
+ "output_tokens": 179,
1137
+ "reasoning_output_tokens": 78
1138
+ },
1139
+ {
1140
+ "cache_write_input_tokens": 0,
1141
+ "cached_input_tokens": 38784,
1142
+ "input_tokens": 39356,
1143
+ "output_tokens": 795,
1144
+ "reasoning_output_tokens": 487
1145
+ },
1146
+ {
1147
+ "cache_write_input_tokens": 0,
1148
+ "cached_input_tokens": 40064,
1149
+ "input_tokens": 40560,
1150
+ "output_tokens": 412,
1151
+ "reasoning_output_tokens": 214
1152
+ },
1153
+ {
1154
+ "cache_write_input_tokens": 0,
1155
+ "cached_input_tokens": 40832,
1156
+ "input_tokens": 41227,
1157
+ "output_tokens": 230,
1158
+ "reasoning_output_tokens": 30
1159
+ },
1160
+ {
1161
+ "cache_write_input_tokens": 0,
1162
+ "cached_input_tokens": 41344,
1163
+ "input_tokens": 41999,
1164
+ "output_tokens": 52,
1165
+ "reasoning_output_tokens": 0
1166
+ },
1167
+ {
1168
+ "cache_write_input_tokens": 0,
1169
+ "cached_input_tokens": 41856,
1170
+ "input_tokens": 42422,
1171
+ "output_tokens": 88,
1172
+ "reasoning_output_tokens": 0
1173
+ },
1174
+ {
1175
+ "cache_write_input_tokens": 0,
1176
+ "cached_input_tokens": 42368,
1177
+ "input_tokens": 43056,
1178
+ "output_tokens": 130,
1179
+ "reasoning_output_tokens": 0
1180
+ },
1181
+ {
1182
+ "cache_write_input_tokens": 0,
1183
+ "cached_input_tokens": 43136,
1184
+ "input_tokens": 43618,
1185
+ "output_tokens": 51,
1186
+ "reasoning_output_tokens": 0
1187
+ },
1188
+ {
1189
+ "cache_write_input_tokens": 0,
1190
+ "cached_input_tokens": 43520,
1191
+ "input_tokens": 44040,
1192
+ "output_tokens": 128,
1193
+ "reasoning_output_tokens": 30
1194
+ },
1195
+ {
1196
+ "cache_write_input_tokens": 0,
1197
+ "cached_input_tokens": 44032,
1198
+ "input_tokens": 44597,
1199
+ "output_tokens": 169,
1200
+ "reasoning_output_tokens": 0
1201
+ },
1202
+ {
1203
+ "cache_write_input_tokens": 0,
1204
+ "cached_input_tokens": 44672,
1205
+ "input_tokens": 45306,
1206
+ "output_tokens": 52,
1207
+ "reasoning_output_tokens": 0
1208
+ },
1209
+ {
1210
+ "cache_write_input_tokens": 0,
1211
+ "cached_input_tokens": 45184,
1212
+ "input_tokens": 45725,
1213
+ "output_tokens": 87,
1214
+ "reasoning_output_tokens": 0
1215
+ },
1216
+ {
1217
+ "cache_write_input_tokens": 0,
1218
+ "cached_input_tokens": 45696,
1219
+ "input_tokens": 46351,
1220
+ "output_tokens": 132,
1221
+ "reasoning_output_tokens": 0
1222
+ },
1223
+ {
1224
+ "cache_write_input_tokens": 0,
1225
+ "cached_input_tokens": 46464,
1226
+ "input_tokens": 46912,
1227
+ "output_tokens": 53,
1228
+ "reasoning_output_tokens": 0
1229
+ },
1230
+ {
1231
+ "cache_write_input_tokens": 0,
1232
+ "cached_input_tokens": 46848,
1233
+ "input_tokens": 47103,
1234
+ "output_tokens": 93,
1235
+ "reasoning_output_tokens": 42
1236
+ },
1237
+ {
1238
+ "cache_write_input_tokens": 0,
1239
+ "cached_input_tokens": 47104,
1240
+ "input_tokens": 47579,
1241
+ "output_tokens": 695,
1242
+ "reasoning_output_tokens": 230
1243
+ },
1244
+ {
1245
+ "cache_write_input_tokens": 0,
1246
+ "cached_input_tokens": 48128,
1247
+ "input_tokens": 48301,
1248
+ "output_tokens": 829,
1249
+ "reasoning_output_tokens": 0
1250
+ },
1251
+ {
1252
+ "cache_write_input_tokens": 0,
1253
+ "cached_input_tokens": 49024,
1254
+ "input_tokens": 49157,
1255
+ "output_tokens": 33,
1256
+ "reasoning_output_tokens": 0
1257
+ }
1258
+ ],
1259
+ "usage_source": "provider-raw"
1260
+ },
1261
+ "call_activity": {
1262
+ "schema": "rlebench/call-activity/1",
1263
+ "source": "Kinex native trace",
1264
+ "model_tool_calls": 59,
1265
+ "model_tool_calls_by_name": {
1266
+ "ls": 3,
1267
+ "read": 16,
1268
+ "find": 1,
1269
+ "exec": 31,
1270
+ "devices": 2,
1271
+ "device": 1,
1272
+ "write": 2,
1273
+ "device_stop": 1,
1274
+ "edit": 2
1275
+ },
1276
+ "python_instrumented_model_tool_calls": null,
1277
+ "python_tool_invocations": null,
1278
+ "nested_python_tool_invocations": null,
1279
+ "python_tool_invocations_by_origin": null,
1280
+ "python_device_rpc_attempts": null,
1281
+ "python_device_rpc_attempts_by_action": null,
1282
+ "python_device_rpc_errors": null,
1283
+ "python_scope": "Recorded instrumented Python invocations only; other paths and automatic fatal stops are not counted."
1284
+ },
1285
+ "model_response_count": 49,
1286
+ "usage_source": "Harbor session metrics",
1287
+ "execution": {
1288
+ "status": "finished",
1289
+ "reason": null
1290
+ },
1291
+ "native": {
1292
+ "evidence_valid": true,
1293
+ "success": true,
1294
+ "termination": "success",
1295
+ "steps": 1354
1296
+ },
1297
+ "failure_attribution": {
1298
+ "classification": "native_success",
1299
+ "physical_cause": null,
1300
+ "physical_cause_evidence": null
1301
+ },
1302
+ "exception": null
1303
+ },
1304
+ "model_call_count": 0
1305
+ }
experiments/robotwin-native/kinex-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-3/episode-report.json ADDED
@@ -0,0 +1,462 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "episode": 1,
3
+ "initialization": {
4
+ "expert_plan_success": true,
5
+ "requested_seed": 0
6
+ },
7
+ "native_reward": 1.0,
8
+ "planned_episodes": 1,
9
+ "reason": "success",
10
+ "simulation_time_s": 54.160000000000004,
11
+ "step_limit": 1700,
12
+ "steps": 1354,
13
+ "success": true,
14
+ "valid": true,
15
+ "step": "episode_01",
16
+ "evidence_valid": true,
17
+ "wall_time_sec": 756.194745,
18
+ "usage": {
19
+ "n_input_tokens": 1644067,
20
+ "n_output_tokens": 10285,
21
+ "n_cache_tokens": 1598336,
22
+ "cost_usd": null,
23
+ "n_fresh_input_tokens": 45731,
24
+ "n_cache_read_tokens": 1598336,
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": 0.9721842236356547,
32
+ "cache_read_input_tokens": 1598336,
33
+ "cache_reported_input_tokens": 1644067,
34
+ "cache_write_basis": "provider-reported",
35
+ "cache_write_input_tokens": 0,
36
+ "cache_write_reported_input_tokens": 1644067,
37
+ "cached_input_tokens": 1598336,
38
+ "cost_usd": null,
39
+ "fresh_input_tokens": 45731,
40
+ "harbor_mismatches": {},
41
+ "input_breakdown_complete": true,
42
+ "input_tokens": 1644067,
43
+ "known_cache_write_input_tokens": 0,
44
+ "known_cached_input_tokens": 1598336,
45
+ "known_fresh_input_tokens": 45731,
46
+ "known_input_tokens": 1644067,
47
+ "known_output_tokens": 10285,
48
+ "known_reasoning_output_tokens": 2439,
49
+ "output_tokens": 10285,
50
+ "parser": "harbor/0.22.0",
51
+ "reasoning_output_tokens": 2439,
52
+ "reasoning_reported_output_tokens": 10285,
53
+ "rejected_attempts": 0,
54
+ "reported_responses": {
55
+ "cache_reported_input_tokens": 49,
56
+ "cache_write_input_tokens": 49,
57
+ "cache_write_reported_input_tokens": 49,
58
+ "cached_input_tokens": 49,
59
+ "fresh_input_tokens": 49,
60
+ "input_tokens": 49,
61
+ "output_tokens": 49,
62
+ "reasoning_output_tokens": 49,
63
+ "reasoning_reported_output_tokens": 49
64
+ },
65
+ "request_attempts": 49,
66
+ "response_count": 49,
67
+ "response_ids_complete": true,
68
+ "schema": "rlebench/token-usage/1",
69
+ "source": "Harbor session metrics",
70
+ "total_tokens": 1654352,
71
+ "uncached_input_tokens": 45731,
72
+ "unresolved_attempts": [],
73
+ "usage_records": [
74
+ {
75
+ "cache_write_input_tokens": 0,
76
+ "cached_input_tokens": 0,
77
+ "input_tokens": 11304,
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": 11344,
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": 11477,
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": 14222,
99
+ "output_tokens": 164,
100
+ "reasoning_output_tokens": 0
101
+ },
102
+ {
103
+ "cache_write_input_tokens": 0,
104
+ "cached_input_tokens": 14208,
105
+ "input_tokens": 17281,
106
+ "output_tokens": 173,
107
+ "reasoning_output_tokens": 0
108
+ },
109
+ {
110
+ "cache_write_input_tokens": 0,
111
+ "cached_input_tokens": 17280,
112
+ "input_tokens": 20636,
113
+ "output_tokens": 54,
114
+ "reasoning_output_tokens": 0
115
+ },
116
+ {
117
+ "cache_write_input_tokens": 0,
118
+ "cached_input_tokens": 20608,
119
+ "input_tokens": 21876,
120
+ "output_tokens": 98,
121
+ "reasoning_output_tokens": 0
122
+ },
123
+ {
124
+ "cache_write_input_tokens": 0,
125
+ "cached_input_tokens": 21888,
126
+ "input_tokens": 22170,
127
+ "output_tokens": 39,
128
+ "reasoning_output_tokens": 0
129
+ },
130
+ {
131
+ "cache_write_input_tokens": 0,
132
+ "cached_input_tokens": 22016,
133
+ "input_tokens": 22678,
134
+ "output_tokens": 23,
135
+ "reasoning_output_tokens": 0
136
+ },
137
+ {
138
+ "cache_write_input_tokens": 0,
139
+ "cached_input_tokens": 22528,
140
+ "input_tokens": 22870,
141
+ "output_tokens": 152,
142
+ "reasoning_output_tokens": 25
143
+ },
144
+ {
145
+ "cache_write_input_tokens": 0,
146
+ "cached_input_tokens": 22912,
147
+ "input_tokens": 24389,
148
+ "output_tokens": 362,
149
+ "reasoning_output_tokens": 0
150
+ },
151
+ {
152
+ "cache_write_input_tokens": 0,
153
+ "cached_input_tokens": 24576,
154
+ "input_tokens": 25147,
155
+ "output_tokens": 39,
156
+ "reasoning_output_tokens": 0
157
+ },
158
+ {
159
+ "cache_write_input_tokens": 0,
160
+ "cached_input_tokens": 25088,
161
+ "input_tokens": 25418,
162
+ "output_tokens": 354,
163
+ "reasoning_output_tokens": 0
164
+ },
165
+ {
166
+ "cache_write_input_tokens": 0,
167
+ "cached_input_tokens": 25600,
168
+ "input_tokens": 27068,
169
+ "output_tokens": 321,
170
+ "reasoning_output_tokens": 247
171
+ },
172
+ {
173
+ "cache_write_input_tokens": 0,
174
+ "cached_input_tokens": 27264,
175
+ "input_tokens": 28568,
176
+ "output_tokens": 1068,
177
+ "reasoning_output_tokens": 62
178
+ },
179
+ {
180
+ "cache_write_input_tokens": 0,
181
+ "cached_input_tokens": 29440,
182
+ "input_tokens": 29657,
183
+ "output_tokens": 74,
184
+ "reasoning_output_tokens": 0
185
+ },
186
+ {
187
+ "cache_write_input_tokens": 0,
188
+ "cached_input_tokens": 29568,
189
+ "input_tokens": 30414,
190
+ "output_tokens": 280,
191
+ "reasoning_output_tokens": 43
192
+ },
193
+ {
194
+ "cache_write_input_tokens": 0,
195
+ "cached_input_tokens": 30592,
196
+ "input_tokens": 30842,
197
+ "output_tokens": 252,
198
+ "reasoning_output_tokens": 0
199
+ },
200
+ {
201
+ "cache_write_input_tokens": 0,
202
+ "cached_input_tokens": 30976,
203
+ "input_tokens": 31389,
204
+ "output_tokens": 393,
205
+ "reasoning_output_tokens": 115
206
+ },
207
+ {
208
+ "cache_write_input_tokens": 0,
209
+ "cached_input_tokens": 31616,
210
+ "input_tokens": 32053,
211
+ "output_tokens": 189,
212
+ "reasoning_output_tokens": 94
213
+ },
214
+ {
215
+ "cache_write_input_tokens": 0,
216
+ "cached_input_tokens": 31872,
217
+ "input_tokens": 32718,
218
+ "output_tokens": 51,
219
+ "reasoning_output_tokens": 11
220
+ },
221
+ {
222
+ "cache_write_input_tokens": 0,
223
+ "cached_input_tokens": 32640,
224
+ "input_tokens": 32791,
225
+ "output_tokens": 277,
226
+ "reasoning_output_tokens": 46
227
+ },
228
+ {
229
+ "cache_write_input_tokens": 0,
230
+ "cached_input_tokens": 32896,
231
+ "input_tokens": 33530,
232
+ "output_tokens": 328,
233
+ "reasoning_output_tokens": 126
234
+ },
235
+ {
236
+ "cache_write_input_tokens": 0,
237
+ "cached_input_tokens": 33664,
238
+ "input_tokens": 34081,
239
+ "output_tokens": 283,
240
+ "reasoning_output_tokens": 142
241
+ },
242
+ {
243
+ "cache_write_input_tokens": 0,
244
+ "cached_input_tokens": 34176,
245
+ "input_tokens": 34391,
246
+ "output_tokens": 99,
247
+ "reasoning_output_tokens": 0
248
+ },
249
+ {
250
+ "cache_write_input_tokens": 0,
251
+ "cached_input_tokens": 34304,
252
+ "input_tokens": 35033,
253
+ "output_tokens": 134,
254
+ "reasoning_output_tokens": 0
255
+ },
256
+ {
257
+ "cache_write_input_tokens": 0,
258
+ "cached_input_tokens": 35072,
259
+ "input_tokens": 35705,
260
+ "output_tokens": 52,
261
+ "reasoning_output_tokens": 0
262
+ },
263
+ {
264
+ "cache_write_input_tokens": 0,
265
+ "cached_input_tokens": 35584,
266
+ "input_tokens": 36128,
267
+ "output_tokens": 88,
268
+ "reasoning_output_tokens": 0
269
+ },
270
+ {
271
+ "cache_write_input_tokens": 0,
272
+ "cached_input_tokens": 36096,
273
+ "input_tokens": 36756,
274
+ "output_tokens": 445,
275
+ "reasoning_output_tokens": 353
276
+ },
277
+ {
278
+ "cache_write_input_tokens": 0,
279
+ "cached_input_tokens": 36992,
280
+ "input_tokens": 37742,
281
+ "output_tokens": 155,
282
+ "reasoning_output_tokens": 64
283
+ },
284
+ {
285
+ "cache_write_input_tokens": 0,
286
+ "cached_input_tokens": 37760,
287
+ "input_tokens": 38332,
288
+ "output_tokens": 52,
289
+ "reasoning_output_tokens": 0
290
+ },
291
+ {
292
+ "cache_write_input_tokens": 0,
293
+ "cached_input_tokens": 38272,
294
+ "input_tokens": 38748,
295
+ "output_tokens": 179,
296
+ "reasoning_output_tokens": 78
297
+ },
298
+ {
299
+ "cache_write_input_tokens": 0,
300
+ "cached_input_tokens": 38784,
301
+ "input_tokens": 39356,
302
+ "output_tokens": 795,
303
+ "reasoning_output_tokens": 487
304
+ },
305
+ {
306
+ "cache_write_input_tokens": 0,
307
+ "cached_input_tokens": 40064,
308
+ "input_tokens": 40560,
309
+ "output_tokens": 412,
310
+ "reasoning_output_tokens": 214
311
+ },
312
+ {
313
+ "cache_write_input_tokens": 0,
314
+ "cached_input_tokens": 40832,
315
+ "input_tokens": 41227,
316
+ "output_tokens": 230,
317
+ "reasoning_output_tokens": 30
318
+ },
319
+ {
320
+ "cache_write_input_tokens": 0,
321
+ "cached_input_tokens": 41344,
322
+ "input_tokens": 41999,
323
+ "output_tokens": 52,
324
+ "reasoning_output_tokens": 0
325
+ },
326
+ {
327
+ "cache_write_input_tokens": 0,
328
+ "cached_input_tokens": 41856,
329
+ "input_tokens": 42422,
330
+ "output_tokens": 88,
331
+ "reasoning_output_tokens": 0
332
+ },
333
+ {
334
+ "cache_write_input_tokens": 0,
335
+ "cached_input_tokens": 42368,
336
+ "input_tokens": 43056,
337
+ "output_tokens": 130,
338
+ "reasoning_output_tokens": 0
339
+ },
340
+ {
341
+ "cache_write_input_tokens": 0,
342
+ "cached_input_tokens": 43136,
343
+ "input_tokens": 43618,
344
+ "output_tokens": 51,
345
+ "reasoning_output_tokens": 0
346
+ },
347
+ {
348
+ "cache_write_input_tokens": 0,
349
+ "cached_input_tokens": 43520,
350
+ "input_tokens": 44040,
351
+ "output_tokens": 128,
352
+ "reasoning_output_tokens": 30
353
+ },
354
+ {
355
+ "cache_write_input_tokens": 0,
356
+ "cached_input_tokens": 44032,
357
+ "input_tokens": 44597,
358
+ "output_tokens": 169,
359
+ "reasoning_output_tokens": 0
360
+ },
361
+ {
362
+ "cache_write_input_tokens": 0,
363
+ "cached_input_tokens": 44672,
364
+ "input_tokens": 45306,
365
+ "output_tokens": 52,
366
+ "reasoning_output_tokens": 0
367
+ },
368
+ {
369
+ "cache_write_input_tokens": 0,
370
+ "cached_input_tokens": 45184,
371
+ "input_tokens": 45725,
372
+ "output_tokens": 87,
373
+ "reasoning_output_tokens": 0
374
+ },
375
+ {
376
+ "cache_write_input_tokens": 0,
377
+ "cached_input_tokens": 45696,
378
+ "input_tokens": 46351,
379
+ "output_tokens": 132,
380
+ "reasoning_output_tokens": 0
381
+ },
382
+ {
383
+ "cache_write_input_tokens": 0,
384
+ "cached_input_tokens": 46464,
385
+ "input_tokens": 46912,
386
+ "output_tokens": 53,
387
+ "reasoning_output_tokens": 0
388
+ },
389
+ {
390
+ "cache_write_input_tokens": 0,
391
+ "cached_input_tokens": 46848,
392
+ "input_tokens": 47103,
393
+ "output_tokens": 93,
394
+ "reasoning_output_tokens": 42
395
+ },
396
+ {
397
+ "cache_write_input_tokens": 0,
398
+ "cached_input_tokens": 47104,
399
+ "input_tokens": 47579,
400
+ "output_tokens": 695,
401
+ "reasoning_output_tokens": 230
402
+ },
403
+ {
404
+ "cache_write_input_tokens": 0,
405
+ "cached_input_tokens": 48128,
406
+ "input_tokens": 48301,
407
+ "output_tokens": 829,
408
+ "reasoning_output_tokens": 0
409
+ },
410
+ {
411
+ "cache_write_input_tokens": 0,
412
+ "cached_input_tokens": 49024,
413
+ "input_tokens": 49157,
414
+ "output_tokens": 33,
415
+ "reasoning_output_tokens": 0
416
+ }
417
+ ],
418
+ "usage_source": "provider-raw"
419
+ },
420
+ "call_activity": {
421
+ "schema": "rlebench/call-activity/1",
422
+ "source": "Kinex native trace",
423
+ "model_tool_calls": 59,
424
+ "model_tool_calls_by_name": {
425
+ "ls": 3,
426
+ "read": 16,
427
+ "find": 1,
428
+ "exec": 31,
429
+ "devices": 2,
430
+ "device": 1,
431
+ "write": 2,
432
+ "device_stop": 1,
433
+ "edit": 2
434
+ },
435
+ "python_instrumented_model_tool_calls": null,
436
+ "python_tool_invocations": null,
437
+ "nested_python_tool_invocations": null,
438
+ "python_tool_invocations_by_origin": null,
439
+ "python_device_rpc_attempts": null,
440
+ "python_device_rpc_attempts_by_action": null,
441
+ "python_device_rpc_errors": null,
442
+ "python_scope": "Recorded instrumented Python invocations only; other paths and automatic fatal stops are not counted."
443
+ },
444
+ "model_response_count": 49,
445
+ "usage_source": "Harbor session metrics",
446
+ "execution": {
447
+ "status": "finished",
448
+ "reason": null
449
+ },
450
+ "native": {
451
+ "evidence_valid": true,
452
+ "success": true,
453
+ "termination": "success",
454
+ "steps": 1354
455
+ },
456
+ "failure_attribution": {
457
+ "classification": "native_success",
458
+ "physical_cause": null,
459
+ "physical_cause_evidence": null
460
+ },
461
+ "exception": null
462
+ }
experiments/robotwin-native/kinex-3/episode.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "episode": 1,
3
+ "initialization": {
4
+ "expert_plan_success": true,
5
+ "requested_seed": 0
6
+ },
7
+ "native_reward": 1.0,
8
+ "planned_episodes": 1,
9
+ "reason": "success",
10
+ "simulation_time_s": 54.160000000000004,
11
+ "step_limit": 1700,
12
+ "steps": 1354,
13
+ "success": true,
14
+ "valid": true
15
+ }
experiments/robotwin-native/kinex-3/evidence.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d3884a8c19a136e796d8a761058c2dcb24b467aef261ab6e3872c8b91f4c487
3
+ size 12619391
experiments/robotwin-native/kinex-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-3/media.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "native_frames": 543,
3
+ "published_frames": 543,
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": "c3ff59536ab0",
18
+ "task": "robotwin/put-bottles-dustbin-aloha-agilex",
19
+ "state": "complete",
20
+ "views": {
21
+ "third_person": {
22
+ "received": 543,
23
+ "dropped": 0,
24
+ "segments": [
25
+ {
26
+ "file": "third_person.mp4",
27
+ "frames": 543,
28
+ "width": 960,
29
+ "height": 720,
30
+ "fps": 10.0,
31
+ "first_stamp_ns": 0,
32
+ "last_stamp_ns": 54160000000
33
+ }
34
+ ],
35
+ "error": null
36
+ },
37
+ "left_wrist": {
38
+ "received": 543,
39
+ "dropped": 0,
40
+ "segments": [
41
+ {
42
+ "file": "left_wrist.mp4",
43
+ "frames": 543,
44
+ "width": 960,
45
+ "height": 720,
46
+ "fps": 10.0,
47
+ "first_stamp_ns": 0,
48
+ "last_stamp_ns": 54160000000
49
+ }
50
+ ],
51
+ "error": null
52
+ },
53
+ "right_wrist": {
54
+ "received": 543,
55
+ "dropped": 0,
56
+ "segments": [
57
+ {
58
+ "file": "right_wrist.mp4",
59
+ "frames": 543,
60
+ "width": 960,
61
+ "height": 720,
62
+ "fps": 10.0,
63
+ "first_stamp_ns": 0,
64
+ "last_stamp_ns": 54160000000
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": 543,
84
+ "mean_ssim": 0.9985112486187845,
85
+ "minimum_ssim": 0.996403,
86
+ "label_area_excluded": true
87
+ },
88
+ "left_wrist": {
89
+ "frames_compared": 543,
90
+ "mean_ssim": 0.9969153756906078,
91
+ "minimum_ssim": 0.990548,
92
+ "label_area_excluded": true
93
+ },
94
+ "right_wrist": {
95
+ "frames_compared": 543,
96
+ "mean_ssim": 0.9981290165745856,
97
+ "minimum_ssim": 0.996466,
98
+ "label_area_excluded": true
99
+ }
100
+ }
101
+ }
experiments/robotwin-native/kinex-3/observations/1c00b1c6-1.png ADDED
experiments/robotwin-native/kinex-3/observations/1c00b1c6-2.png ADDED
experiments/robotwin-native/kinex-3/observations/25fb6783-1.png ADDED
experiments/robotwin-native/kinex-3/observations/2e34d545-1.png ADDED
experiments/robotwin-native/kinex-3/observations/38f9e622-1.png ADDED
experiments/robotwin-native/kinex-3/observations/38f9e622-2.png ADDED
experiments/robotwin-native/kinex-3/observations/40cd7359-1.png ADDED
experiments/robotwin-native/kinex-3/observations/4a031853-1.png ADDED
experiments/robotwin-native/kinex-3/observations/4a031853-2.png ADDED
experiments/robotwin-native/kinex-3/observations/5306339c-1.png ADDED
experiments/robotwin-native/kinex-3/observations/5306339c-2.png ADDED
experiments/robotwin-native/kinex-3/observations/59f4e1b6-1.png ADDED
experiments/robotwin-native/kinex-3/observations/7af7cb21-1.png ADDED
experiments/robotwin-native/kinex-3/observations/82847509-1.png ADDED
experiments/robotwin-native/kinex-3/observations/8c8cb210-1.png ADDED
experiments/robotwin-native/kinex-3/observations/8c8cb210-2.png ADDED
experiments/robotwin-native/kinex-3/observations/9deb50d9-1.png ADDED
experiments/robotwin-native/kinex-3/observations/9deb50d9-2.png ADDED
experiments/robotwin-native/kinex-3/observations/a2501cc5-1.png ADDED
experiments/robotwin-native/kinex-3/observations/a57d6a5b-1.png ADDED
experiments/robotwin-native/kinex-3/observations/b27b5e13-1.png ADDED
experiments/robotwin-native/kinex-3/observations/bd40c7f2-1.png ADDED
experiments/robotwin-native/kinex-3/observations/bd40c7f2-2.png ADDED
experiments/robotwin-native/kinex-3/observations/e0aba462-1.png ADDED
experiments/robotwin-native/kinex-3/observations/e8dcb044-1.png ADDED
experiments/robotwin-native/kinex-3/observations/e8dcb044-2.png ADDED
experiments/robotwin-native/kinex-3/observations/ed3b0775-1.png ADDED
experiments/robotwin-native/kinex-3/observations/ed3b0775-2.png ADDED
experiments/robotwin-native/kinex-3/observations/eda38a00-1.png ADDED
experiments/robotwin-native/kinex-3/observations/eda38a00-2.png ADDED
experiments/robotwin-native/kinex-3/observations/ff2d32db-1.png ADDED
experiments/robotwin-native/kinex-3/policy-calls.json ADDED
@@ -0,0 +1 @@
 
 
1
+ []
experiments/robotwin-native/kinex-3/poster.jpg ADDED
experiments/robotwin-native/kinex-3/provider-requests.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
experiments/robotwin-native/kinex-3/session.json ADDED
The diff for this file is too large to render. See raw diff
 
experiments/robotwin-native/kinex-3/session.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
experiments/robotwin-native/kinex-3/terminal.jpg ADDED
experiments/robotwin-native/kinex-3/tool-calls.json ADDED
The diff for this file is too large to render. See raw diff