Gallery · LIBERO-Pro

place the bowl on top of the cabinet

libero_goal_lan, task 4 · seed 3. DeepSeek-V4.1-Flash, code mode: 6 decisions, 37 motions, solved.

2× · 65 s of motion
  1. 1 of 6 · grasp_bowl.py
  2. 2 of 6 · grasp2.py
  3. 3 of 6 · grasp3.py
  4. 4 of 6 · grasp4.py
  5. 5 of 6 · place1.py
  6. 6 of 6 · place3.py
  1. move_pose
  2. move_pose
  3. set_gripper
  4. wait_ticks
  5. move_pose
  6. wait_ticks
  7. set_gripper
  8. move_pose
  9. move_pose
  10. move_pose
  11. wait_ticks
  12. set_gripper
  13. wait_ticks
  14. set_gripper
  15. move_pose
  16. wait_ticks
  17. set_gripper
  18. wait_ticks
  19. move_pose
  20. move_pose
  21. wait_ticks
  22. move_pose
  23. wait_ticks
  24. move_pose
  25. wait_ticks
  26. set_gripper
  27. wait_ticks
  28. set_gripper
  29. move_pose
  30. wait_ticks
  31. move_pose
  32. wait_ticks
  33. move_pose
  34. wait_ticks
  35. move_pose
  36. wait_ticks
  37. move_pose
Decisions

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.