Gallery · LIBERO-Pro

put the left moka pot on the stove

libero_10_task, task 8 · seed 0. DeepSeek-V4.1-Flash, code mode: 12 decisions, 36 motions, solved.

2× · 70 s of motion
  1. 1 of 12 · step1.py
  2. 2 of 12 · step2.py
  3. 3 of 12 · step3.py
  4. 4 of 12 · step4.py
  5. 5 of 12 · step5.py
  6. 6 of 12 · step6.py
  7. 7 of 12 · step6.py
  8. 8 of 12 · step7.py
  9. 9 of 12 · step7b.py
  10. 10 of 12 · step7c.py
  11. 11 of 12 · step8.py
  12. 12 of 12 · step9.py
  1. set_gripper
  2. move_pose
  3. move_pose
  4. set_gripper
  5. move_pose
  6. move_pose
  7. move_pose
  8. move_pose
  9. set_gripper
  10. move_pose
  11. wait_ticks
  12. set_gripper
  13. move_pose
  14. move_pose
  15. move_pose
  16. move_pose
  17. set_gripper
  18. move_pose
  19. move_pose
  20. set_gripper
  21. move_pose
  22. move_pose
  23. move_pose
  24. move_pose
  25. move_pose
  26. set_gripper
  27. move_pose
  28. wait_ticks
  29. set_gripper
  30. move_pose
  31. move_pose
  32. set_gripper
  33. move_pose
  34. move_pose
  35. move_pose
  36. 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.