List your company

Adversarial and Jailbreaking Attack Detection on MaliciousInstruct

0.8825AUROCOpenAI Omni
0.1145640.3139320.51330.712668Feb 4, 2026

Evaluation results

MethodLinks
OpenAI OmniMode=Zero-shot2026.020.88250.92
LLAMAGUARD4-12BMode=Zero-shot2026.020.87181
WILDGUARDMode=Zero-shot2026.020.86170.9982
FDBDMode=Zero-shot2026.020.8510.71
LLAMAGUARD3-1BMode=Zero-shot2026.020.85071
POLYGUARDMode=Zero-shot2026.020.85010.78
T3-GMMMode=Zero-shot2026.020.8280.59
MDJUDGEMode=Zero-shot2026.020.79570.97
DUOGUARDMode=Zero-shot2026.020.77451
T3-OCSVMMode=Zero-shot2026.020.75860.68
Perspective APIMode=Zero-shot2026.020.68281
OPENMAXMode=Zero-shot2026.020.59541
LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.020.57910.93
RMDMode=Zero-shot2026.020.38691
ADASCALEMode=Zero-shot2026.020.29941
CIDERMode=Zero-shot2026.020.2581
GMMMode=Zero-shot2026.020.22521
REACTMode=Zero-shot2026.020.19131
NNGUIDEMode=Zero-shot2026.020.16221
VIMMode=Zero-shot2026.020.14411