List your company

Adversarial and Jailbreaking Attack Detection on BeaverTails

0.8525AUROCDUOGUARD
0.1914760.3630880.53470.706312Feb 4, 2026

Evaluation results

MethodLinks
DUOGUARDMode=Zero-shot2026.020.85250.8064
WILDGUARDMode=Zero-shot2026.020.82181
OpenAI OmniMode=Zero-shot2026.020.80910.9488
POLYGUARDMode=Zero-shot2026.020.80710.7269
Perspective APIMode=Zero-shot2026.020.79220.8429
MDJUDGEMode=Zero-shot2026.020.77790.8987
T3-GMMMode=Zero-shot2026.020.72760.7847
FDBDMode=Zero-shot2026.020.72260.8754
LLAMAGUARD4-12BMode=Zero-shot2026.020.71371
LLAMAGUARD3-1BMode=Zero-shot2026.020.71351
OPENMAXMode=Zero-shot2026.020.62260.9984
T3-OCSVMMode=Zero-shot2026.020.60810.8758
LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.020.55980.9366
RMDMode=Zero-shot2026.020.35680.9727
ADASCALEMode=Zero-shot2026.020.28330.9986
CIDERMode=Zero-shot2026.020.27770.9974
REACTMode=Zero-shot2026.020.26550.998
GMMMode=Zero-shot2026.020.25150.9972
NNGUIDEMode=Zero-shot2026.020.22390.9994
VIMMode=Zero-shot2026.020.21690.9997