Adversarial and Jailbreaking Attack Detection on BeaverTails
0.8525AUROCDUOGUARD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DUOGUARDMode=Zero-shot2026.02 | 0.8525 | 0.8064 | |
| WILDGUARDMode=Zero-shot2026.02 | 0.8218 | 1 | |
| OpenAI OmniMode=Zero-shot2026.02 | 0.8091 | 0.9488 | |
| POLYGUARDMode=Zero-shot2026.02 | 0.8071 | 0.7269 | |
| Perspective APIMode=Zero-shot2026.02 | 0.7922 | 0.8429 | |
| MDJUDGEMode=Zero-shot2026.02 | 0.7779 | 0.8987 | |
| T3-GMMMode=Zero-shot2026.02 | 0.7276 | 0.7847 | |
| FDBDMode=Zero-shot2026.02 | 0.7226 | 0.8754 | |
| LLAMAGUARD4-12BMode=Zero-shot2026.02 | 0.7137 | 1 | |
| LLAMAGUARD3-1BMode=Zero-shot2026.02 | 0.7135 | 1 | |
| OPENMAXMode=Zero-shot2026.02 | 0.6226 | 0.9984 | |
| T3-OCSVMMode=Zero-shot2026.02 | 0.6081 | 0.8758 | |
| LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.02 | 0.5598 | 0.9366 | |
| RMDMode=Zero-shot2026.02 | 0.3568 | 0.9727 | |
| ADASCALEMode=Zero-shot2026.02 | 0.2833 | 0.9986 | |
| CIDERMode=Zero-shot2026.02 | 0.2777 | 0.9974 | |
| REACTMode=Zero-shot2026.02 | 0.2655 | 0.998 | |
| GMMMode=Zero-shot2026.02 | 0.2515 | 0.9972 | |
| NNGUIDEMode=Zero-shot2026.02 | 0.2239 | 0.9994 | |
| VIMMode=Zero-shot2026.02 | 0.2169 | 0.9997 |