List your company

Jailbreak Attack on Harmful behaviors jailbreak evaluation set

100ASR (Harmful Behaviors)Baseline
-4235077May 27, 2026

Evaluation results

MethodLinks
BaselineTarget Model=Qwen3-8B2026.05100
BaselineTarget Model=Qwen3-1.7B2026.05100
BaselineTarget Model=Phi-4-reasoning2026.0597
BaselineTarget Model=GPT-OSS-20B2026.0594
BaselineTarget Model=Gemma-4-E4B-it2026.0588
BCTTarget Model=Gemma-4-E4B-it2026.0557
BCTTarget Model=Qwen3-1.7B2026.0536
ACTTarget Model=Gemma-4-E4B-it2026.0533
BCTTarget Model=Qwen3-8B2026.0530
BCTTarget Model=Phi-4-reasoning2026.0515.3
ACTTarget Model=Qwen3-1.7B2026.0514
ACTTarget Model=GPT-OSS-20B2026.051
BCTTarget Model=GPT-OSS-20B2026.051
ACTTarget Model=Qwen3-8B2026.050
ACTTarget Model=Phi-4-reasoning2026.050