Gallery · LIBERO-Pro

slide open the top drawer and place the bowl in it

libero_goal_lan, task 3 · seed 2. DeepSeek-V4.1-Flash, code mode: 12 decisions, 32 motions, solved.

2× · 107 s of motion
  1. 1 of 12 · step1.py
  2. 2 of 12 · open_drawer.py
  3. 3 of 12 · after_open.py
  4. 4 of 12 · after_open2.py
  5. 5 of 12 · go_bowl.py
  6. 6 of 12 · grasp_bowl.py
  7. 7 of 12 · survey_tray.py
  8. 8 of 12 · grasp_bowl2.py
  9. 9 of 12 · grasp_bowl3.py
  10. 10 of 12 · grasp_bowl4.py
  11. 11 of 12 · carry_bowl.py
  12. 12 of 12 · release3.py
  1. move_to
  2. move_pose
  3. move_pose
  4. move_pose
  5. set_gripper
  6. move_pose
  7. set_gripper
  8. move_pose
  9. set_gripper
  10. set_gripper
  11. move_pose
  12. move_to
  13. move_to
  14. move_pose
  15. move_pose
  16. move_pose
  17. set_gripper
  18. set_gripper
  19. move_pose
  20. move_pose
  21. move_pose
  22. move_pose
  23. move_pose
  24. set_gripper
  25. set_gripper
  26. move_to
  27. set_gripper
  28. move_pose
  29. set_gripper
  30. move_pose
  31. set_gripper
  32. set_gripper
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.