Robot Manipulation on FRBench-Real Adversarial
60Pour Water Success RateRePO-VLA (Phase I)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RePO-VLA (Phase I)Training Phase=Phase I, Trials per task=10, Adversarial=true2026.05 | 60 | 30 | 40 | 20 | 37.5 | — | |
| RePO-VLA (Full 1x)Training Phase=Full, Data Scaling=1x, Trials per task=10, Adversarial=true2026.05 | 30 | 30 | 30 | 30 | 30 | 10 | |
| π0Venue=RSS'25, Trials per task=10, Adversarial=true2026.05 | 20 | 0 | 10 | 20 | 12.5 | — | |
| π0.5Venue=CoRL'25, Trials per task=10, Adversarial=true2026.05 | 20 | 10 | 30 | 20 | 20 | — |