Failure Detection on WidowX seen
0.096Brier ScoreSAFE-RNN-TDQC (Ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| SAFE-RNN-TDQC (Ours)VLA Model=OpenVLA2026.04 | 0.096 | |
| SAFE-MLP BCEVLA Model=OpenVLA2026.04 | 0.127 | |
| SAFE-MLP-TDQC (Ours)VLA Model=OpenVLA2026.04 | 0.13 | |
| RNN-TDQC (Ours)VLA Model=OpenVLA2026.04 | 0.156 | |
| SAFE-RNNVLA Model=OpenVLA2026.04 | 0.169 | |
| Running Avg prob.VLA Model=OpenVLA2026.04 | 0.255 | |
| Avg prob.VLA Model=OpenVLA2026.04 | 0.275 | |
| RNN-BCEVLA Model=OpenVLA2026.04 | 0.301 | |
| Avg entropyVLA Model=OpenVLA2026.04 | 0.414 | |
| Running Avg entropyVLA Model=OpenVLA2026.04 | 0.435 | |
| Max prob.VLA Model=OpenVLA2026.04 | 0.572 |