{ "schema": "efe-task-dag/1", "taxonomy": "v2", "taskKey": "robopaint/acrylic_applesoranges", "suite": "robopaint", "base": "robopaint-acrylic-applesoranges-i00", "title_zh": "丙烯临摹苹果与橙子", "summary_zh": "先使干刷载上颜料,再在主纸上以不超过390笔且符合路径和扫笔规则的丙烯笔触临摹目标静物画,使整体颜色、结构、边缘和绘画覆盖均达到评分阈值。", "embodiment_zh": "两种模式均为固定底座的七自由度 Franka Panda 单臂,作用末端是安装工具(圆形丙烯画刷,15 mm软刷毛,无夹爪),在桌面主纸与左侧十色颜料井之间操作;底座不移动。特权模式在本地仿真生成并提交轨迹,标准模式经服务控制同一机械臂。", "task_class": "compositional", "analysis_level": "step", "spec": { "version": "1.2", "task": "robopaint-acrylic-applesoranges-i00", "bench": "robopaint", "progress": "live", "entities": {}, "milestones": [ { "id": "brush_loaded", "zh": "笔刷蘸颜料", "mode": "event", "when": null, "after": [], "hold_s": 0, "freeze_held": false, "goal": false, "criterion_zh": "机械臂末端圆刷的笔尖进入任一颜料井圆内(距井心小于 12 mm),并降至颜料液面 z=0.028 m 以下,使干刷装载该井颜料;场景视角可见刷尖蘸入左侧颜料盘,刷毛带上颜色,随后主纸出现有色笔迹也可证明已装载。仅悬停在井上方、空刷碰纸且不留色均未完成;装载一次即记录完成,后续换色不另计步骤。", "step": { "cat": "其他", "item": "蘸料" }, "contact": [ { "who": "手持工具", "how": "其他", "plan": 1, "note_zh": "候选接触方式:浸蘸。Panda 单臂末端安装圆刷的刷毛进入颜料井接触颜料;这是刷毛载料,现有抓握、压/擦、托/舀均不能准确描述浸蘸。" } ], "optional": false, "verify_zh": null, "sources": [ "inputs/task/privileged/instruction.md:10-15", "inputs/task/privileged/environment/instance.json:68-80", "inputs/task/privileged/environment/instance.json:213", "inputs/grader/efe_robopaint-files/rcb_brush.py:432-442", "inputs/grader/efe_robopaint-files/rcb_brush.py:274-279", "inputs/grader/efe_robopaint-files/rcb_brush.py:875-909" ], "conf": 6, "note_zh": "初态是干刷,未装载时绘画压力为零,所有合法临摹路线都须先载料。按 S2 / D11 保留一次必经前置,不加工具拿起节点;蘸料不是舀起液体,也不是在纸上画写,步骤类型暂用其他候选“蘸料”。" }, { "id": "still_life_copied", "zh": "临摹静物画", "mode": "state", "when": null, "after": [ "brush_loaded" ], "hold_s": 0, "freeze_held": false, "goal": true, "criterion_zh": "录像右下角的主纸俯视图应重现目标整幅画:棕色背景、上中部碗盘与黄褐水果、左中红褐苹果及下部白灰桌布褶皱的色块和位置相符,大片留白或只有少量轮廓不算完成;继续落笔覆盖造成图像指标失达标即失去该状态。评分先作每轴至多 5 mm 的统一整像素平移(最优位置在搜索边界时不平移;不旋转、不缩放),按 1 mm 高斯模糊、绘画并集外扩 2 mm 区域计算平均色差 ΔE≤2、90分位 ΔE≤6、全图亮度 SSIM≥0.92(1.5 mm 窗口)、区域亮度梯度相关≥0.8、覆盖率≥0.5的绘画区域 IoU≥0.95,并要求主纸有色连续接触≤390笔、着色接触总路径≤33900 mm,终末 scribble_share≤0.1(每笔路径长度/其包围盒对角线与10 mm的较大值,比例>2.5者的路径占总路径之比);练习纸和干刷接触不计,完成不另要求抬笔或静置。", "step": { "cat": "材料", "item": "画/写" }, "contact": [ { "who": "手持工具", "how": "压/擦", "plan": 1, "note_zh": "Panda 单臂末端安装的圆形丙烯画刷以刷毛压在主纸上扫出颜料笔触,刷毛与主纸接触;工具安装在臂上,不重复记握刷接触。" } ], "optional": false, "verify_zh": null, "sources": [ "inputs/task/privileged/instruction.md:27-61", "inputs/task/standard/instruction.md:41-67", "inputs/grader/efe_robopaint-pictures-files/acrylic-applesoranges-i00/target.png", "inputs/task/privileged/environment/instance.json:215-230", "inputs/grader/efe_robopaint-files/rcb_brush_score.py:365-446", "inputs/grader/efe_robopaint-files/rcb_brush_score.py:454-465", "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:140-202", "inputs/oracle/replay.mp4 @ 0.5, 7.5, 22.5, 42.5, 59.0 s", "inputs/oracle/run.json: reward" ], "conf": 6, "note_zh": "一幅画为一个聚合结果,颜色区域、水果数量、不同指标和最多390笔都是验收,不据此拆节点或添加绘画顺序。主纸可在刷子仍接触时达标,故不冻结持物状态;步骤按当前画面和截至当前的过程统计评价,scribble_share必须在终末达标,途中暂时超限可被后续合规路径稀释,不设历史否决。" } ], "constraints": [ { "id": "stroke_cap_exceeded", "zh": "主纸笔数超限", "never": null, "criterion_zh": "同一次受评分绘画中,主纸上载有颜料的连续接触笔数超过390;练习纸和干刷接触不计。超出后无法靠覆盖旧画减少已记笔数,最终 stroke_count_ok 为假并否决成功,不提前终止。", "consequence": "历史否决", "sources": [ "inputs/task/privileged/environment/instance.json:226", "inputs/grader/efe_robopaint-files/rcb_brush_score.py:440-458" ] }, { "id": "paint_path_exceeded", "zh": "着色路径超限", "never": null, "criterion_zh": "同一次受评分绘画中,所有主纸有色接触笔触的刷尖平面路径总长超过33900 mm(按100 Hz物理样本累计);练习纸不计。该累计量不可减少,path_budget_ok 为假,最终 strict_ok 和 success 被否决,不提前终止。", "consequence": "历史否决", "sources": [ "inputs/task/privileged/environment/instance.json:227", "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:70-91", "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:145-160", "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:185-202" ] } ], "success": null, "atoms": [ { "atom": "brush_loaded", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush.py:875-884; log painter.loaded and painter.color after each control step" }, { "atom": "de_mean", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:365-439; log score.de_mean" }, { "atom": "de_p90", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:436; log score.de_p90" }, { "atom": "ssim_L", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:437; log score.ssim_L" }, { "atom": "edge_corr", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:421-425; log score.edge_corr" }, { "atom": "paint_iou", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:409-412,438; log score.paint_iou" }, { "atom": "n_strokes", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_score.py:440-446; log score.n_strokes including current painted contact" }, { "atom": "path_mm", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:70-91,145-160; log process.path_mm" }, { "atom": "scribble_share", "level": "new_code", "source": "inputs/grader/efe_robopaint-files/rcb_brush_strict.py:171-179; log process.scribble_share" } ] }, "native_metrics": [ { "name": "de_mean", "unit": "deltaE_ab", "threshold": 2, "direction": "lower_better" }, { "name": "de_p90", "unit": "deltaE_ab", "threshold": 6, "direction": "lower_better" }, { "name": "ssim_L", "unit": "ratio", "threshold": 0.92, "direction": "higher_better" }, { "name": "edge_corr", "unit": "correlation", "threshold": 0.8, "direction": "higher_better" }, { "name": "paint_iou", "unit": "ratio", "threshold": 0.95, "direction": "higher_better" }, { "name": "n_strokes", "unit": "count", "threshold": 390, "direction": "lower_better" }, { "name": "path_mm", "unit": "mm", "threshold": 33900, "direction": "lower_better" }, { "name": "scribble_share", "unit": "ratio", "threshold": 0.1, "direction": "lower_better" } ], "open_questions_zh": [ "inputs/states/index.json 的 logs 为空,未提供任何逐帧状态日志;按 P8 所有节点 when、约束 never 和 success 均为 null。评分器虽能计算这些指标,仍不能假定它们已经记录;atoms 列出未来需补记的字段,图像数值阈值无法仅凭低分辨率录像精确复算。", "S2 / D11、CASES.md OPEN-21:干刷必先蘸料,本文件保留一次必经载料步骤;用户允许省略此节点。按文字与初态强制前置列表先定为 compositional;蘸料的步骤与接触类型仍缺准确词,暂用其他候选“蘸料”和“浸蘸”,不强归为舀或画写。", "文本步骤未受评:评分器未把蘸料次数或最终 loaded 单独列为成功合取;它是干刷开始且仅装载后能着色的物理前置,而非最终目标。", "两模式的画面和过程规则相同,但总裁决不同:特权模式需所有新进程重放成功且最终 state_sha 一致、实例匹配(privileged/grade.py:122-125);标准模式对服务实时画布评分,并要求实时裁决有效且无仿真错误(standard/grade.py:29-39;sim/server.py:85-118)。这些交付、重放有效性和基础设施条件不建语义步骤;无状态日志,success 无法表达总裁决。两次相同轨迹重放不是两个独立语义场景。", "D11 / D6:本任务复制一幅图像,按整画一个结果处理;390是最多笔数,不是必须完成的重复次数,不按水果、颜色连通区或参考笔触拆节点,也不要求某一颜色先画。", "过程禁令由评分器的路径总长和终末 scribble_share 两个数落实,没有额外的蛇形或轮廓填充分型判定;scribble_share 不是单调累计量,途中超0.1不能直接判历史否决。", "最大接触倾角、最大压深、接触速度、笔柄接触样本和蘸料次数等只作诊断记录,没有独立成功阈值,依 S13 不加约束;文字要求竖直持刷是绘画方式,不另拆步骤。", "N2:唯一成功参考 run 为 pht8xwh,也在被标注集合中,故无独立成功参考可替换;已查看其59.7秒录像及 reward,未从其换色顺序或386笔策略推导依赖。其记录的7200秒会话预算与当前说明3600秒不同,不改变共同绘画判据。" ], "conf": 6, "conf_reason_zh": "语义步骤、干刷初态、必要依赖和全部数值阈值有文字与评分代码依据,成功录像末帧及原生指标支持整画结果;无逐帧状态日志,无法绑定或检验条件,且蘸料类型仍属开放边界,因此取6分。", "review": { "status": "agreed", "rounds": 1, "reviewers": [ { "engine": "claude", "model": "claude-sonnet-5-5", "round": 1, "verdict": "agree", "conf": 7, "summary_zh": "节点(蘸料前置 + 临摹整幅画)、目标集合、390笔与33900 mm约束、终末scribble_share放入验收均与评分器一致;蘸料步骤按D11/S2可保留,无状态日志故when/never/success为null符合P8。无需修改内容。", "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:['brush_loaded', 'still_life_copied']" ] }, "provenance": { "prompt": "prompts/01_task_dag.md", "prompt_sha256": "b8e79c84b73950529a4fa6b988412b2d221d5df4e052adafcd9605afac64aa25", "engine": "codex", "model": "gpt-6.1-sol", "effort": "high", "created": "2026-10-29T10:36:06Z", "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/robopaint__acrylic_applesoranges/01_dag" }, { "stage": "dag_review", "round": 1, "ok": true, "space": "work/tasks/robopaint__acrylic_applesoranges/02_review_r1" } ], "review_engine": "claude", "review_model": "claude-sonnet-5-5", "inputs_manifest_sha256": "95e706540faae5bb31220d946e384bb1c089fa87835f49622f4a03b8502a9f07", "repo_commit": "80cac7a5a227db930c659c3307a1e79c2a84bc53" } }