Adversarial and Jailbreaking Attack Detection on JailbreakBench
0.8622AUROCT3-OCSVM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| T3-OCSVMMode=Zero-shot2026.02 | 0.8622 | 0.4539 | |
| OpenAI OmniMode=Zero-shot2026.02 | 0.8369 | 0.6724 | |
| POLYGUARDMode=Zero-shot2026.02 | 0.7904 | 0.7201 | |
| T3-GMMMode=Zero-shot2026.02 | 0.7588 | 0.5358 | |
| MDJUDGEMode=Zero-shot2026.02 | 0.7302 | 0.8908 | |
| LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.02 | 0.7293 | 0.6689 | |
| LLAMAGUARD3-1BMode=Zero-shot2026.02 | 0.7248 | 1 | |
| Perspective APIMode=Zero-shot2026.02 | 0.7233 | 0.9795 | |
| LLAMAGUARD4-12BMode=Zero-shot2026.02 | 0.7165 | 1 | |
| WILDGUARDMode=Zero-shot2026.02 | 0.6978 | 1 | |
| DUOGUARDMode=Zero-shot2026.02 | 0.682 | 0.9898 | |
| FDBDMode=Zero-shot2026.02 | 0.6195 | 0.9317 | |
| CIDERMode=Zero-shot2026.02 | 0.5966 | 0.9556 | |
| OPENMAXMode=Zero-shot2026.02 | 0.5861 | 0.9898 | |
| GMMMode=Zero-shot2026.02 | 0.5377 | 0.9625 | |
| RMDMode=Zero-shot2026.02 | 0.4816 | 0.9898 | |
| VIMMode=Zero-shot2026.02 | 0.4032 | 1 | |
| NNGUIDEMode=Zero-shot2026.02 | 0.3795 | 0.9863 | |
| ADASCALEMode=Zero-shot2026.02 | 0.35 | 0.9829 | |
| REACTMode=Zero-shot2026.02 | 0.257 | 1 |