Multilingual Blending: LLM Safety Alignment Evaluation with Language Mixture

Jiayang Song, Yuheng Huang, Zhehua Zhou +1

As safety remains a crucial concern throughout the development lifecycle of Large Language Models (LLMs), researchers and industrial practitioners have increasingly focused on safe…