{ "schema": "efe-task-dag/1", "taxonomy": "v2", "taskKey": "bigym2/sandwich_remove", "suite": "bigym2", "base": "bigym2-sandwich-remove-i00", "title_zh": "三明治移上砧板", "summary_zh": "三明治以任一面朝上、偏离水平不超过10度地接触砧板,连续保持50个控制步,且全程手指未碰三明治、机器人未摔倒。", "embodiment_zh": "三个模式均为Unitree G1双足人形机器人,双臂各装二指夹爪;腿部和腰部由全身控制器驱动,可行走、下蹲和前倾,无轮式底盘。", "task_class": "atomic", "analysis_level": "step", "spec": { "version": "1.2", "task": "bigym2-sandwich-remove-i00", "bench": "bigym2", "progress": "live", "entities": {}, "milestones": [ { "id": "sandwich_on_board", "zh": "三明治上砧板", "mode": "state", "when": null, "after": [], "hold_s": 0.98, "freeze_held": false, "goal": true, "criterion_zh": "双外部视角可见金黄色三明治已从黑色圆煎锅移到长方形木砧板上并与板接触,任一面朝上均可,自身法向距世界向上或向下不超过10度,连续50个控制步成立(名义1秒,采样首尾间隔0.98秒)。仍在锅中、悬在板上或板外、立起或倾斜超限均未完成;离板或失去平放状态即失去该结果,全程手指不得碰三明治且机器人不得摔倒。", "step": { "cat": "关系", "item": "放置" }, "contact": [ { "who": "手持工具", "how": "托/舀", "plan": 1, "note_zh": "一侧夹爪持锅铲,铲面从下方托住三明治,送到砧板上;夹爪握铲柄属于同一工具链,不重复标接触。" }, { "who": "腿/躯干", "how": "支撑", "plan": 1, "note_zh": "双腿和足部与地面接触,在锅边与右侧砧板之间调整站位,支承持铲搬移三明治的机器人。" }, { "who": "手持工具", "how": "托/舀", "plan": 2, "note_zh": "一侧夹爪握煎锅锅柄,锅底托住三明治,移到砧板处再倾锅使其落到板上;成功录像用左夹爪,随后抬离锅,不限定用手。" }, { "who": "腿/躯干", "how": "支撑", "plan": 2, "note_zh": "双腿和足部接触地面并侧步,支承持锅搬移三明治的机器人。" } ], "optional": false, "verify_zh": null, "sources": [ "inputs/task/privileged/instruction.md:3-19", "inputs/task/standard/instruction.md:7-32", "inputs/task/standard-realtime/instruction.md:8-32", "inputs/grader/efe_bigym2-files/privileged/environment/rcb_bigym2.py:12-16,147-157", "inputs/oracle/run.json:1 steps[2] (id=4), out.text: upstream envs/pick_and_place.py:344-353", "inputs/oracle/run.json:1 steps[34] (id=36), calls: final trajectory generator; reward.success=1", "inputs/oracle/replay.mp4:0.00-22.76s", "docs/DECISIONS.md:D31 (50-1)*0.02=0.98 s" ], "conf": 6, "note_zh": "一个搬移结果,不拆出锅、抓握、抬起、搬运、松手或等待;平放、接触与保持窗口均为本步验收,不额外要求静止速度、居中或锅铲撤离。初态三明治在敞口锅内、砧板为空,无必开机构;取锅铲并非所有受评成功必经。抓着辅助工具也可能达到结果,因此关闭持物冻结。" } ], "constraints": [ { "id": "finger_touches_sandwich", "zh": "手指碰三明治", "never": null, "criterion_zh": "任一夹爪手指直接碰到三明治,按任务文字立即失败;上游父类使用is_gripper_holding_object检查,具体接触实现未随输入提供。", "consequence": "立即终止", "sources": [ "inputs/task/privileged/instruction.md:15", "inputs/oracle/run.json:1 steps[3] (id=5), out.text: ToastSandwich._fail" ] }, { "id": "robot_fell", "zh": "机器人摔倒", "never": null, "criterion_zh": "机器人在任意时刻被上游记为摔倒:任务文字给出骨盆相对竖直倾斜超过约53度,或骨盆高度低于0.30米;即使后来站起、三明治已平放砧板,成功仍被否决,摔倒本身不立即结束回合。", "consequence": "历史否决", "sources": [ "inputs/task/privileged/instruction.md:15", "inputs/grader/efe_bigym2-files/privileged/environment/rcb_bigym2.py:14-16,147-157" ] }, { "id": "pelvis_out_of_bounds", "zh": "骨盆越界", "never": null, "criterion_zh": "骨盆距场景原点超过10米,上游立即结束回合;距离采用何种范数及坐标分量需上游源码核实,不自行替换成平面距离。", "consequence": "立即终止", "sources": [ "inputs/task/privileged/instruction.md:17-19", "inputs/grader/efe_bigym2-files/privileged/environment/rcb_bigym2.py:14-16" ] } ], "success": null, "atoms": [ { "atom": "sandwich_board_contact", "level": "new_code", "source": "需回放适配器记录sandwich.is_colliding(board);inputs/oracle/run.json:1 steps[2]摘录上游envs/pick_and_place.py:351。" }, { "atom": "sandwich_level_angle_deg", "level": "new_code", "source": "需记录三明治法向与世界向上及向下夹角的较小值;inputs/oracle/run.json:1 steps[2]摘录上游envs/pick_and_place.py:345-349;10度阈值见inputs/task/privileged/instruction.md:15。" }, { "atom": "sandwich_goal_hold_steps", "level": "new_code", "source": "需记录上游连续成功控制步计数,或密集记录接触和平放状态重建50步窗口;inputs/task/privileged/instruction.md:15-19。" }, { "atom": "finger_sandwich_violation", "level": "new_code", "source": "需记录ToastSandwich._fail使用的逐夹爪is_gripper_holding_object(sandwich, side)及其接触判据;inputs/oracle/run.json:1 steps[3];不能使用未经核实的通用held替代。" }, { "atom": "robot_ever_fell", "level": "new_code", "source": "需记录env.episode_fell()的历史值;inputs/grader/efe_bigym2-files/privileged/environment/rcb_bigym2.py:156-157。" }, { "atom": "pelvis_out_of_bounds", "level": "new_code", "source": "需记录上游10米骨盆越界检查;inputs/task/privileged/instruction.md:17-19;尚无对应状态字段。" }, { "atom": "benchmark_success", "level": "new_code", "source": "需逐帧记录bigym.loco.eval.is_success(env),而非交付评分;inputs/grader/efe_bigym2-files/privileged/environment/rcb_bigym2.py:147-150。" } ] }, "native_metrics": [ { "name": "sandwich_level_angle_deg", "unit": "deg", "threshold": 10, "direction": "lower_better" } ], "open_questions_zh": [ "输入inputs/states/index.json的logs为空,不能安全绑定任何现有状态字段;依P8,节点when、约束never和success均为null,原子列出需补充的记录,录像可判定终态但无法逐控制步重核50步窗口。", "任务设计问题:文本指定用锅铲,但oracle在不使用锅铲的情况下抓锅柄翻锅倒出三明治,reward.success=1;可见上游RemoveSandwich._success摘录只检验平放和砧板接触。依S3、D2及CASES.md STEP-7,不加入必需锅铲拿起节点,保留两种工具接触方案。", "依D5,将离开煎锅并移上砧板解释为一个放置步骤:文本只要求从锅到板的搬移,出锅是达成板上关系的中间过程,没有独立出锅目标。", "输入评分器委托bigym上游包,但没有完整上游任务模块、基础失败检查、保持计数与摔倒检测源码;仅oracle运行记录含部分源码摘录。10度、50步、约53度、0.30米和10米取自instruction,未估计新阈值;手指触碰与is_gripper_holding_object的精确对应、越界距离定义及父类是否另有失败条件均待核实。", "三个模式共享Episode及上游is_success,初态哈希一致;privileged另要求两个回放均成功且最终状态一致,standard与standard-realtime直接采用现场回合判定。重复验证及基础设施有效性不构成世界状态节点;本文件无日志可表达这些交付层判据,success保留null。" ], "conf": 6, "conf_reason_zh": "一个放置步骤直接来自三个模式的文本,初态和成功录像支持其粒度、结果与替代工具方案;成功及失败阈值有任务文字和部分上游摘录依据,但没有状态日志,完整上游接触、计数和基础失败实现不可核查。", "review": { "status": "agreed", "rounds": 1, "reviewers": [ { "engine": "claude", "model": "claude-sonnet-5-5", "round": 1, "verdict": "agree", "conf": 7, "summary_zh": "单节点(三明治上砧板)符合文本与STEP-7/Y17下的开放读法:锅铲未列为节点并写入open_questions_zh,任务设计问题已标注;约束三条与评分器/文本一致;无状态日志故when=null、success=null合理。", "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:['sandwich_on_board']" ] }, "provenance": { "prompt": "prompts/01_task_dag.md", "prompt_sha256": "1142cdb56758062bbd0ff585ad99f3fefe07a96377b61614dc4f50b2d16600b7", "engine": "codex", "model": "gpt-6.1-sol", "effort": "high", "created": "2026-10-29T01:41:24Z", "review_prompt": "prompts/02_task_dag_review.md", "review_prompt_sha256": "65812ffc3cb33aaeb10dbe28c925ffe5535a356f8132ec431dcf58b59c2a0b85", "reconcile_prompt": "prompts/05_reconcile.md", "reconcile_prompt_sha256": "b3cd3803cf22f21db996e7ab1145811f823e92c7728c6f21d3b6649cc1ed0e83", "calls": [ { "stage": "dag", "round": 0, "ok": true, "space": "work/tasks/bigym2__sandwich_remove/01_dag" }, { "stage": "dag_review", "round": 1, "ok": true, "space": "work/tasks/bigym2__sandwich_remove/02_review_r1" } ], "review_engine": "claude", "review_model": "claude-sonnet-5-5", "inputs_manifest_sha256": "d6b2e0b3668b3a5a9b1bc73f073bcb8a0a3bafd40a914badf55ab65ecc280571", "repo_commit": "4c6d01b9990e3b78ddb620bbd237932b3696f60e" } }