Real-to-sim Robot Manipulation on SimplerEnv Bridge
95.8Spoon Success RateX-VLA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| X-VLAMaximum inference horizon=120 steps2026.05 | 95.8 | 75 | 62.5 | 70.8 | 76 | |
| GTA-VLA2026.05 | 95.8 | 87.5 | 66.7 | 75 | 81.2 | |
| Uni-VLA2026.05 | 83.3 | 66.7 | 33.3 | 95.8 | 69.8 | |
| GR00T-N12026.05 | 64.5 | 65.5 | 5.5 | 93 | 57.1 | |
| π02026.05 | 50 | 41.7 | 29.2 | 70.8 | 47.9 | |
| ThinkAct2026.05 | 37.5 | 8.7 | 58.3 | 70.8 | 43.8 | |
| SpatialVLA*Status=fine-tuned, Reproduced=true2026.05 | 25 | 12.5 | 16.7 | 66.6 | 30.2 | |
| SpatialVLA* + OursStatus=fine-tuned, Reproduced=true2026.05 | 25 | 12.5 | 20.8 | 75 | 33.3 | |
| RoboVLMsStatus=pre-trained2026.05 | 20.8 | 25 | 8.3 | 0 | 13.5 | |
| RoboVLMsStatus=fine-tuned2026.05 | 20.8 | 25 | 12.5 | 58.3 | 31.3 | |
| SpatialVLAStatus=pre-trained2026.05 | 12.3 | 20.8 | 20.8 | 54.2 | 27 | |
| OpenVLA2026.05 | 4.2 | 0 | 8.3 | 45.8 | 14.6 | |
| OpenVLA2026.05 | 0 | 0 | 4.1 | 0 | 1 |