List your company

Adversarial and Jailbreaking Attack Detection on JailbreakBench

0.8622AUROCT3-OCSVM
0.2327920.3961960.55960.723004Feb 4, 2026

Evaluation results

MethodLinks
T3-OCSVMMode=Zero-shot2026.020.86220.4539
OpenAI OmniMode=Zero-shot2026.020.83690.6724
POLYGUARDMode=Zero-shot2026.020.79040.7201
T3-GMMMode=Zero-shot2026.020.75880.5358
MDJUDGEMode=Zero-shot2026.020.73020.8908
LLAMAGUARD3-1B-LOGITSMode=Zero-shot2026.020.72930.6689
LLAMAGUARD3-1BMode=Zero-shot2026.020.72481
Perspective APIMode=Zero-shot2026.020.72330.9795
LLAMAGUARD4-12BMode=Zero-shot2026.020.71651
WILDGUARDMode=Zero-shot2026.020.69781
DUOGUARDMode=Zero-shot2026.020.6820.9898
FDBDMode=Zero-shot2026.020.61950.9317
CIDERMode=Zero-shot2026.020.59660.9556
OPENMAXMode=Zero-shot2026.020.58610.9898
GMMMode=Zero-shot2026.020.53770.9625
RMDMode=Zero-shot2026.020.48160.9898
VIMMode=Zero-shot2026.020.40321
NNGUIDEMode=Zero-shot2026.020.37950.9863
ADASCALEMode=Zero-shot2026.020.350.9829
REACTMode=Zero-shot2026.020.2571