pour by language, layout 3
Task: “Pending official task instruction from environment reset.” DeepSeek-V4.1-Flash, code mode: 10 decisions, 69 motions, after step limit. The benchmark's check passed after its official step limit.
2× · 50 s of motion
- 1 of 10 · step_grasp.py
- 2 of 10 · step_pour2.py
- 3 of 10 · step_setdown.py
- 4 of 10 · s1_grasp.py
- 5 of 10 · s2_pour.py
- 6 of 10 · s3_setdown.py
- 7 of 10 · s1_grasp.py
- 8 of 10 · s2_pour.py
- 9 of 10 · s2b_continue.py
- 10 of 10 · s3_setdown.py
- move_pose
- move_pose
- set_gripper
- move_pose
- set_gripper
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- move_pose
- wait_ticks
- move_pose
- move_pose
- set_gripper
- move_pose
- move_pose
- move_pose
- set_gripper
- move_pose
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- move_pose
- wait_ticks
- move_pose
- move_pose
- set_gripper
- move_pose
- move_pose
- move_pose
- set_gripper
- move_pose
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- wait_ticks
- move_pose
- wait_ticks
- move_pose
Decisions
pour_by_language layout 3: Code mode · 10 decisions, solved after step limit
The recording is the robot's own cameras at 2×; the time between motions, while the model thinks, is not filmed. The 3D replay is the camera's picture in depth after each decision, with the paths it planned, to turn and step through.