Visual Planning on RoboDesk VP2 benchmark
43.33Upright Block Success RateCLAW
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CLAW2026.06 | 43.33 | — | 25.56 | 11.11 | 12.22 | — | |
| WARM8K+E2E30K2025.10 | 35.33 | 4 | 19.33 | 18 | 29.33 | 21.2 | |
| LAM30K+WM30K2025.10 | 22 | 2.67 | 3.33 | 4.67 | 6 | 7.73 | |
| LAPO2026.06 | 15.56 | — | 3.33 | 6.67 | 8.89 | — | |
| ADAWORLD2025.10 | 5 | 5.83 | 10 | 10.83 | 29.17 | 12.17 | |
| AdaWorldsource=Reported from original paper2026.06 | 5 | — | 10 | 10.83 | 29.17 | — | |
| Random2026.06 | 0 | — | 1.11 | 1.11 | 10 | — |