List your company

Jailbreak Attack on GuidedBench

29.5ASR (HarmBench)Original Weight
1.428.711623.29Apr 2, 2025

Evaluation results

MethodLinks
Original WeightModel=Mistral 7B Instruct v0.22025.0429.523.73
Original WeightModel=Llama3 8B Instruct2025.0467.92
CB*Model=Mistral 7B Instruct v0.22025.044.56.77
REPBENDModel=Mistral 7B Instruct v0.22025.043.56.72
REPBENDModel=Llama3 8B Instruct2025.0436.75
CB*Model=Llama3 8B Instruct2025.042.57.5