List your company

Adversarial and Jailbreaking Attack Detection on AdvBench

0.9675AUROCT3-GMM
0.2459480.4332740.62060.807926Feb 4, 2026

Evaluation results

MethodLinks
T3-GMMMode=Zero-shot2026.020.96750.1577
T3-OCSVMMode=Zero-shot2026.020.95780.1731
OpenAI OmniMode=Zero-shot2026.020.89080.8269
LLAMAGUARD3-1BMode=Zero-shot2026.020.88941
LLAMAGUARD4-12BMode=Zero-shot2026.020.88221
POLYGUARDMode=Zero-shot2026.020.8670.6654
WILDGUARDMode=Zero-shot2026.020.86581
DUOGUARDMode=Zero-shot2026.020.82410.9327
LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.020.8110.75
Perspective APIMode=Zero-shot2026.020.78950.9558
MDJUDGEMode=Zero-shot2026.020.78140.9942
ADASCALEMode=Zero-shot2026.020.58940.9827
FDBDMode=Zero-shot2026.020.56890.975
REACTMode=Zero-shot2026.020.44610.9962
NNGUIDEMode=Zero-shot2026.020.39890.9981
OPENMAXMode=Zero-shot2026.020.36810.9942
RMDMode=Zero-shot2026.020.35750.9981
VIMMode=Zero-shot2026.020.3340.9981
CIDERMode=Zero-shot2026.020.29631
GMMMode=Zero-shot2026.020.27370.9981