Mobile Manipulation on MobileManiBench Unseen Object & Unseen Scene G1 robot
28.2Success RateMobileManiVLA
Evaluation results
| Method | Links | |
|---|---|---|
| MobileManiVLAModel Size=3B, Input Modality=Head-view RGB-D, wrist-view RGB-D, wrist pose states2026.02 | 28.2 | |
| pi_0.5Model Size=3B, Input Modality=Head-view RGB, wrist-view RGB, wrist pose states2026.02 | 18.8 | |
| pi_0Model Size=3B, Input Modality=Head-view RGB, wrist-view RGB, wrist pose states2026.02 | 11.2 | |
| CogACTModel Size=7B, Input Modality=Head-view RGB image2026.02 | 6.8 | |
| OpenVLAModel Size=7B, Input Modality=Head-view RGB image2026.02 | 4.5 |