Robotic Manipulation on RoboTurk
96.32Success RateFULLSOFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FULLSOFTReasoning Strategy=OpenVLA-family re-evaluation2026.05 | 96.32 | 0.445 | |
| VISUALTHINK-VLAReasoning Strategy=OpenVLA-family re-evaluation2026.05 | 96.1 | 0.415 | |
| BASEVLAReasoning Strategy=OpenVLA-family re-evaluation2026.05 | 95.56 | 0.342 | |
| TraceVLAReasoning Strategy=Spatial or image-grounded thinking2026.05 | 93.27 | 0.416 | |
| SpatialVLAReasoning Strategy=Spatial or image-grounded thinking2026.05 | 91.5 | 0.585 | |
| ECoTReasoning Strategy=Textual CoT / autoregressive reasoning before action2026.05 | 15.78 | 6.397 | |
| InternVLA-M1Reasoning Strategy=Spatial or image-grounded thinking2026.05 | 8.51 | 1.151 | |
| DeepThinkVLA-RLReasoning Strategy=Reinforced or hybrid thinking-action decoding2026.05 | 2.12 | 1.323 |