Robot Manipulation on Real-world post-training dataset Task 2: Move condiment cup into slot 1.0 (test)
86.7Success RateANY3D-VLA
Evaluation results
| Method | Links | |
|---|---|---|
| ANY3D-VLATraining data setting=Setting 2 (hybrid point clouds), Inference point cloud source=DA3 (Depth Anything 3)2026.01 | 86.7 | |
| ANY3D-VLATraining data setting=Setting 2 (hybrid point clouds), Inference point cloud source=RealSense2026.01 | 66.7 | |
| ANY3D-VLATraining data setting=Setting 3 (RealSense point clouds), Inference point cloud source=RealSense2026.01 | 60 | |
| ANY3D-VLATraining data setting=Setting 3 (RealSense point clouds), Inference point cloud source=DA3 (Depth Anything 3)2026.01 | 60 | |
| GraspVLA2026.01 | 53.3 | |
| π0.52026.01 | 26.7 | |
| SpatialVLA2026.01 | 6.7 |