File size: 12,636 Bytes
c19b8cb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
{
 "schema": "efe-task-dag/1",
 "taxonomy": "v2",
 "taskKey": "dexjoco/fold_glasses",
 "suite": "dexjoco",
 "base": "dexjoco-fold-glasses-i00",
 "title_zh": "折合眼镜并入盒",
 "summary_zh": "两条镜腿的铰链角均严格大于 1.1 rad,眼镜中心处于敞口盒中心的指定世界轴范围内,以上条件同时连续维持 50 个控制步。",
 "embodiment_zh": "两种模式均为固定基座的单臂 Franka Panda,右臂末端安装一只16自由度 Allegro 灵巧手;没有移动底盘。",
 "task_class": "compositional",
 "analysis_level": "step",
 "spec": {
  "version": "1.2",
  "task": "dexjoco-fold-glasses-i00",
  "bench": "dexjoco",
  "progress": "live",
  "entities": {},
  "milestones": [
   {
    "mode": "state",
    "when": null,
    "hold_s": 0,
    "goal": true,
    "freeze_held": false,
    "optional": false,
    "verify_zh": null,
    "conf": 6,
    "id": "temples_closed_1",
    "zh": "第1条镜腿折合",
    "after": [],
    "criterion_zh": "至少一条黑色镜腿从向后展开的位置绕铰链向镜框内侧收拢,角度严格大于 1.1 rad;右侧腕部近景可见该镜腿横向靠近或跨过镜片背面,另一条仍可展开,若两条都重新展开至不大于阈值则失去。最终两条镜腿折合与眼镜中心入盒须同时连续成立 50 个控制步(50 Hz)。",
    "step": {
     "cat": "机构",
     "item": "关上"
    },
    "contact": [
     {
      "who": "手指/夹爪",
      "how": "推/扫",
      "note_zh": "右侧 Allegro 灵巧手的手指抵住镜腿侧面,绕镜腿铰链向镜框内侧拨合。",
      "plan": 1
     },
     {
      "who": "手指/夹爪",
      "how": "抓握",
      "note_zh": "右侧 Allegro 灵巧手的拇指与其余手指夹住镜腿,转动镜腿使其向镜框内侧折合。",
      "plan": 2
     }
    ],
    "sources": [
     "inputs/task/privileged/instruction.md:3-5 任务文字、展开初态与两条镜腿",
     "inputs/task/privileged/instruction.md:35-49 两个铰链角均须严格大于 1.1 rad",
     "inputs/grader/efe_dexjoco-files/privileged/environment/rcb_dexjoco.py:189-190 对象及铰链映射",
     "inputs/oracle/replay.mp4 @ 0, 3, 6, 9, 12, 15, 18, 21 s 镜腿折合的可见过程",
     "inputs/oracle/run.json:1 steps[id=2].out.text 两个初始铰链角均为 0"
    ],
    "note_zh": "两条镜腿受同一阈值检查且允许任一先合,采用当前折合数量而不固定左右身份;第2条的边仅表示数量由至少一条到两条,不指定动作顺序。镜腿为刚性铰接部件,折合标为机构·关上;持物时也可达标。连续50步为跨三节点的联合成功验收,不分别延迟镜腿进度。"
   },
   {
    "mode": "state",
    "when": null,
    "hold_s": 0,
    "goal": true,
    "freeze_held": false,
    "optional": false,
    "verify_zh": null,
    "conf": 6,
    "id": "temples_closed_2",
    "zh": "第2条镜腿折合",
    "after": [
     "temples_closed_1"
    ],
    "criterion_zh": "两条黑色镜腿都绕各自铰链向镜框内侧收拢,各自角度严格大于 1.1 rad;腕部近景可见两条镜腿靠向镜框背面、可以相互交叉,只收拢一条不算,任一条重新展开至不大于阈值即失去。最终两条镜腿折合与眼镜中心入盒须同时连续成立 50 个控制步(50 Hz)。",
    "step": {
     "cat": "机构",
     "item": "关上"
    },
    "contact": [
     {
      "who": "手指/夹爪",
      "how": "推/扫",
      "note_zh": "右侧 Allegro 灵巧手的手指抵住镜腿侧面,绕镜腿铰链向镜框内侧拨合。",
      "plan": 1
     },
     {
      "who": "手指/夹爪",
      "how": "抓握",
      "note_zh": "右侧 Allegro 灵巧手的拇指与其余手指夹住镜腿,转动镜腿使其向镜框内侧折合。",
      "plan": 2
     }
    ],
    "sources": [
     "inputs/task/privileged/instruction.md:3-5 任务文字、展开初态与两条镜腿",
     "inputs/task/privileged/instruction.md:35-49 两个铰链角均须严格大于 1.1 rad",
     "inputs/grader/efe_dexjoco-files/privileged/environment/rcb_dexjoco.py:189-190 对象及铰链映射",
     "inputs/oracle/replay.mp4 @ 0, 3, 6, 9, 12, 15, 18, 21 s 镜腿折合的可见过程",
     "inputs/oracle/run.json:1 steps[id=2].out.text 两个初始铰链角均为 0"
    ],
    "note_zh": "两条镜腿受同一阈值检查且允许任一先合,采用当前折合数量而不固定左右身份;第2条的边仅表示数量由至少一条到两条,不指定动作顺序。镜腿为刚性铰接部件,折合标为机构·关上;持物时也可达标。连续50步为跨三节点的联合成功验收,不分别延迟镜腿进度。"
   },
   {
    "mode": "state",
    "when": null,
    "hold_s": 0,
    "goal": true,
    "freeze_held": false,
    "optional": false,
    "verify_zh": null,
    "conf": 6,
    "id": "glasses_in_case",
    "zh": "眼镜入盒",
    "after": [],
    "criterion_zh": "左上外部视角中,黑色眼镜移到原来位于其右侧的浅棕色敞口木盒内,并降低到盒壁包围的高度;验收只要求眼镜中心相对盒体模型中心的世界坐标差满足 x、y 各在 [-0.145, 0.145] m,z 在 [-0.045, 0.0275] m,手仍接触眼镜也可成立,停在盒外或仅悬在盒口上方不算,中心离开该范围即失去。最终此位置与两条镜腿角度各严格大于 1.1 rad 必须同时连续成立 50 个 50 Hz 控制步;任一条件中断,联合计数清零。",
    "step": {
     "cat": "关系",
     "item": "放入"
    },
    "contact": [
     {
      "who": "手指/夹爪",
      "how": "抓握",
      "note_zh": "右侧 Allegro 灵巧手的拇指与食指捏住眼镜框及收拢的镜腿,将眼镜送入敞口木盒。",
      "plan": 1
     }
    ],
    "sources": [
     "inputs/task/privileged/instruction.md:3-5 眼镜入已打开的盒",
     "inputs/task/privileged/instruction.md:10-33 以传感器眼镜中心减去盒体模型位置,按世界轴检查中心包围范围",
     "inputs/task/privileged/instruction.md:44-60 角度与位置联合计数,条件中断即清零",
     "inputs/task/standard/instruction.md:4 相同成功条件",
     "inputs/oracle/run.json:1 steps[id=2].out.text 眼镜初始位置 [-0.39026,-0.2144,1.02235],盒中心 [-0.27158,0.28072,1.01235]",
     "inputs/oracle/run.json:1 steps[id=2].out.text _success_trigger_target=50",
     "inputs/oracle/replay.mp4 @ 24, 27, 30, 33.8 s 眼镜搬入木盒并降至盒壁内"
    ],
    "note_zh": "不要求整副眼镜每个部位都被盒体包住,不添加松手、静止、朝向或关盒条件。盒体已打开;评分器没有折合后才能入盒的原生阶段门控,给定盒口也无证据排除先入盒再折合或同步完成,因此不加镜腿到入盒的顺序边。联合保持窗口属于成功验收;无日志时 success=null,由视频判据描述,hold_s=0 保留单步位置达成的即时进度。"
   }
  ],
  "constraints": [],
  "success": null,
  "atoms": [
   {
    "atom": "temple_0_closed",
    "level": "new_code",
    "source": "待记录 data.sensor(\"glass_joint_0_pos\").data[0],或对应 glass_joint_0 的实时关节角;inputs/task/privileged/instruction.md:36-45。条件为角度 > 1.1 rad。"
   },
   {
    "atom": "temple_1_closed",
    "level": "new_code",
    "source": "待记录 data.sensor(\"glass_joint_1_pos\").data[0],或对应 glass_joint_1 的实时关节角;inputs/task/privileged/instruction.md:37-46。条件为角度 > 1.1 rad。"
   },
   {
    "atom": "glasses_center_in_case",
    "level": "new_code",
    "source": "待记录 sensor(\"glass_pos\") 世界坐标与 model.body_pos[model.body(\"open_box\").id],按世界轴相减;inputs/task/privileged/instruction.md:10-33。x/y 范围各 ±0.145 m,z 范围 [-0.045,0.0275] m。"
   },
   {
    "atom": "joint_success_hold",
    "level": "new_code",
    "source": "待记录环境 _success_trigger_count 和 _success_trigger_target;inputs/task/privileged/instruction.md:44-60,inputs/oracle/run.json:1 steps[id=2].out.text 给出 target=50。以计数 >=50 验收联合保持,不需要从视频估计单个关节的精确值。"
   }
  ]
 },
 "native_metrics": [
  {
   "name": "temple_0_angle",
   "unit": "rad",
   "threshold": 1.1,
   "direction": "higher_better"
  },
  {
   "name": "temple_1_angle",
   "unit": "rad",
   "threshold": 1.1,
   "direction": "higher_better"
  },
  {
   "name": "abs_glass_case_dx",
   "unit": "m",
   "threshold": 0.145,
   "direction": "lower_better"
  },
  {
   "name": "abs_glass_case_dy",
   "unit": "m",
   "threshold": 0.145,
   "direction": "lower_better"
  },
  {
   "name": "glass_case_dz_lower_bound",
   "unit": "m",
   "threshold": -0.045,
   "direction": "higher_better"
  },
  {
   "name": "glass_case_dz_upper_bound",
   "unit": "m",
   "threshold": 0.0275,
   "direction": "lower_better"
  }
 ],
 "open_questions_zh": [
  "没有状态日志:inputs/states/index.json 的 logs 为空。依 P8 所有节点 when=null,success=null;精确铰链角、中心边界和联合计数均不能做状态复核,需依据视频和原生成功证据判断,不能把 run.json 的最终评分当成逐帧世界状态绑定。",
  "所给评分链调用 DexJoCo 环境,但未包含 panda_fold_glasses_env.py 完整源码或冻结 instance.npz;成功几何依据 instruction.md 内嵌的原生检查代码,初态由文字、成功视频首帧及 oracle/run.json 第2步实时读数互证。包装器在原生 terminated 且非成功时结束,但输入未完整展示本任务的专有失败条件;不凭空添加掉落或倾倒约束。",
  "指令内嵌代码第59行注释写连续10次,而任务文字与 oracle/run.json 第2步打印的 _success_trigger_target=50 一致,实际按50个控制步接受;50 Hz 下首末采样跨度为0.98秒,联合计数窗口放在整体成功验收,不另设等待节点。",
  "按 N4 将受相同阈值检查的两条镜腿解释为同类计数项,分成第1条、第2条;第2条仅接第1条,不固定 temple_0/temple_1 先后。机构·关上表达铰接镜腿折合;没有柔性材料形变。",
  "两种模式共用 Session 的原生任务成功检查(包装器 step 第425-445行,标准服务 server.py 第60、78行);特权模式额外复播并核对终态哈希,标准模式评判实时单次回合。这些交付及复现检查不增加世界状态节点。"
 ],
 "conf": 6,
 "conf_reason_zh": "三个文字步骤、角度与位置阈值以及联合50步验收均有任务文字和成功视频支持;仅保留计数顺序边。缺少任何状态日志及完整原生环境源码,精确绑定与失败规则无法独立复核。",
 "review": {
  "status": "agreed",
  "rounds": 1,
  "reviewers": [
   {
    "engine": "claude",
    "model": "claude-sonnet-5-5",
    "round": 1,
    "verdict": "agree",
    "conf": 7,
    "summary_zh": "文字步骤(折合两条镜腿、眼镜入盒)与三个节点一致;阈值1.1 rad、盒内范围±0.145/-0.045~0.0275、50步保持均与指令内嵌代码相符;无状态日志故 when=null、success=null 合理;无习惯边(入盒不依赖折合)。未发现需修改之处。",
    "n_issues": 0,
    "n_major": 0
   }
  ],
  "open_zh": []
 },
 "evidence_check": {
  "n_state_runs": 0,
  "n_faithful": 0,
  "success_runs_reaching_all": "0/0",
  "spec_success_eq_grader": "0/0",
  "diverged_runs": [],
  "notes_zh": [
   "when=null 的节点不在状态日志上检查:['temples_closed_1', 'temples_closed_2', 'glasses_in_case']"
  ]
 },
 "provenance": {
  "prompt": "prompts/01_task_dag.md",
  "prompt_sha256": "b8e79c84b73950529a4fa6b988412b2d221d5df4e052adafcd9605afac64aa25",
  "engine": "codex",
  "model": "gpt-6.1-sol",
  "effort": "high",
  "created": "2026-10-29T09:38:10Z",
  "review_prompt": "prompts/02_task_dag_review.md",
  "review_prompt_sha256": "305c2c5feb2340e0c426ebfe883251f74824cbde2c3e230baa9c86fc05c31550",
  "reconcile_prompt": "prompts/05_reconcile.md",
  "reconcile_prompt_sha256": "8c827af3ae3f77d09229399fb41283c50a8f725302cbe3b263b8051213dde2b2",
  "calls": [
   {
    "stage": "dag",
    "round": 0,
    "ok": true,
    "space": "work/tasks/dexjoco__fold_glasses/01_dag"
   },
   {
    "stage": "dag_review",
    "round": 1,
    "ok": true,
    "space": "work/tasks/dexjoco__fold_glasses/02_review_r1"
   }
  ],
  "review_engine": "claude",
  "review_model": "claude-sonnet-5-5",
  "inputs_manifest_sha256": "1a61fc28ae99a99d92c584b523ef825e11bd9d3556a3c7e6f8dfc3eba54c5ff4",
  "repo_commit": "80cac7a5a227db930c659c3307a1e79c2a84bc53"
 }
}