List your company

Attack Detection on Jailbreak Attacks 105K sample set

68Detection RateLogReg (Ours)
27.33637.89348.4559.007Feb 15, 2026

Evaluation results

MethodLinks
LogReg (Ours)input=raw activations, evaluation_regime=LODO2026.0268
Llama-as-Judgeacronym=LJ, prompting=zero-shot2026.0260
PromptGuard 2acronym=PG2026.0248.5
LlamaGuardacronym=LG2026.0228.9