Early Failure Detection on RoboCasa S (Seen)
74.9ROC-AUC (q=0.25)SAFE-LSTM
Evaluation Results
| Method | Links | |
|---|---|---|
| SAFE-LSTMPolicy=π0.52026.06 | 74.9 | |
| SAFE-MLP-TDQCPolicy=π0.52026.06 | 74.8 | |
| SAFE-LSTM-TDQCPolicy=π0.52026.06 | 73.9 | |
| SAFE-MLPPolicy=π0.52026.06 | 73.7 | |
| LogpZOPolicy=π0.52026.06 | 70.7 | |
| ActProbePolicy=π0.52026.06 | 70 | |
| STAC-SinglePolicy=π0.52026.06 | 62.3 | |
| Cosine k-NNPolicy=π0.52026.06 | 56.3 |