List your company

LLM Safety Defense on BeaverTails

10.39ASRMTSA-T3
9.825613.635317.44521.2547May 22, 2025

Evaluation results

MethodLinks
MTSA-T3Target Model=Llama2-7B-Chat, Iteration=32025.0510.39
MTSA-T3Target Model=Zephyr-7B-Beta, Iteration=32025.0511.5
BaselineTarget Model=Llama2-7B-Chat2025.0521.5
BaselineTarget Model=Zephyr-7B-Beta2025.0524.5