Adversarial and Jailbreaking Attack Detection on MaliciousInstruct
0.8825AUROCOpenAI Omni
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenAI OmniMode=Zero-shot2026.02 | 0.8825 | 0.92 | |
| LLAMAGUARD4-12BMode=Zero-shot2026.02 | 0.8718 | 1 | |
| WILDGUARDMode=Zero-shot2026.02 | 0.8617 | 0.9982 | |
| FDBDMode=Zero-shot2026.02 | 0.851 | 0.71 | |
| LLAMAGUARD3-1BMode=Zero-shot2026.02 | 0.8507 | 1 | |
| POLYGUARDMode=Zero-shot2026.02 | 0.8501 | 0.78 | |
| T3-GMMMode=Zero-shot2026.02 | 0.828 | 0.59 | |
| MDJUDGEMode=Zero-shot2026.02 | 0.7957 | 0.97 | |
| DUOGUARDMode=Zero-shot2026.02 | 0.7745 | 1 | |
| T3-OCSVMMode=Zero-shot2026.02 | 0.7586 | 0.68 | |
| Perspective APIMode=Zero-shot2026.02 | 0.6828 | 1 | |
| OPENMAXMode=Zero-shot2026.02 | 0.5954 | 1 | |
| LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.02 | 0.5791 | 0.93 | |
| RMDMode=Zero-shot2026.02 | 0.3869 | 1 | |
| ADASCALEMode=Zero-shot2026.02 | 0.2994 | 1 | |
| CIDERMode=Zero-shot2026.02 | 0.258 | 1 | |
| GMMMode=Zero-shot2026.02 | 0.2252 | 1 | |
| REACTMode=Zero-shot2026.02 | 0.1913 | 1 | |
| NNGUIDEMode=Zero-shot2026.02 | 0.1622 | 1 | |
| VIMMode=Zero-shot2026.02 | 0.1441 | 1 |