Failure Detection on Push Chair (real-world)
100TPRSentinel (STAC + GPT-4o)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Sentinel (STAC + GPT-4o)Det. Time (s)=9.602026.03 | 100 | 90 | 95 | |
| GPT-4o Video QADet. Time (s)=12.892026.03 | 90 | 100 | — | |
| STAC Rev. KLDet. Time (s)=9.832026.03 | 80 | 90 | — | |
| Temporal Non-Distr. Min.Det. Time (s)=9.522026.03 | 70 | 80 | — | |
| Diffusion Output VarianceDet. Time (s)=10.672026.03 | 60 | 90 | — |