Robot Manipulation on Franka Real-world Manipulation Dataset
66Wipe Whiteboard Success RateLift3D-VLA
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Lift3D-VLATraining=Full fine-tuning, Rollouts=15 per task2026.07 | 66 | 66 | 66 | 87 | 93 | 66 | 47 | 73 | 71 | |
| SpatialVLATraining=Full fine-tuning, Rollouts=15 per task2026.07 | 60 | 33 | 20 | 40 | 87 | 33 | 27 | 40 | 43 | |
| π0.5Training=Full fine-tuning, Rollouts=15 per task2026.07 | 60 | 60 | 47 | 87 | 87 | 66 | 53 | 60 | 65 | |
| CoT-VLATraining=Full fine-tuning, Rollouts=15 per task2026.07 | 53 | 66 | 33 | 53 | 47 | 33 | 33 | 53 | 46 |