Failure Detection on LIBERO-PRO Goal-Env
52.8Success RateSampling-based uncertainty
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Sampling-based uncertaintyPolicy Model=π0.5, K=5, Temporal window size (Tw)=52026.06 | 52.8 | 0.736 | 0.681 | |
| Perturbation-based epistemic uncertainty estimationPolicy Model=π0.5, K=5, sigma=0.25, Temporal window size (Tw)=52026.06 | 52.8 | 0.865 | 0.828 |