Contact-rich Robotic Manipulation on Real-world Contact-rich Manipulation Tasks
90Wipe Board Success RateVT-WAM
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| VT-WAMInference Mode=visual-cache, Visual Backbone=Wan2.2-5B, Tactile/Action Experts=1B-scale DiT, Denoising Steps=102026.07 | 90 | 85 | 70 | 81.67 | 60 | 70 | 55 | 61.67 | 71.67 | |
| Fast-WAM2026.07 | 70 | 55 | 45 | 56.67 | 20 | 55 | 25 | 33.33 | 45 | |
| RDP2026.07 | 45 | 60 | 40 | 48.33 | 15 | 35 | 10 | 20 | 34.17 | |
| OmniVTLA2026.07 | 45 | 30 | 25 | 33.33 | 40 | 35 | 40 | 38.33 | 35.83 | |
| π0.52026.07 | 40 | 35 | 35 | 36.67 | 30 | 45 | 10 | 28.33 | 32.5 | |
| DP + TactileDescription=tactile-conditioned diffusion policy2026.07 | 30 | 20 | 25 | 25 | 5 | 35 | 15 | 18.33 | 21.67 |