Robot Manipulation on Real-world Manipulation Tasks In-Domain
100Average Success RateCLASP (auto. perception)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CLASP (auto. perception)Pretraining Data=none†2026.06 | 100 | — | — | — | — | — | — | |
| HarmoWAM2026.05 | 89 | 95 | 90 | 88 | 92 | 85 | 85 | |
| Cosmos-PolicyMethod Category=Joint Modeling WAMs2026.05 | 78 | 93 | 65 | 80 | 83 | 75 | 72 | |
| π0.5Method Category=SOTA VLA policies2026.05 | 74 | 80 | 68 | 75 | 83 | 72 | 67 | |
| VPPMethod Category=Joint Modeling WAMs2026.05 | 73 | 80 | 60 | 78 | 73 | — | — | |
| Wan+AnyposMethod Category=Imagine-then-Execute WAMs2026.05 | 67 | 88 | 60 | 78 | 72 | 53 | 52 | |
| QwenVLA-OFTMethod Category=SOTA VLA policies2026.05 | 63 | 78 | 30 | 73 | 72 | — | — | |
| OpenVLA-OFT+Pretraining Data=20–300 demos2026.06 | 51 | — | — | — | — | — | — | |
| LBMPretraining Data=>10k demos2026.06 | 50 | — | — | — | — | — | — |