2025
Multilingual Blending: Large Language Model Safety Alignment Evaluation with Language Mixture
NAACL 2025findings
As safety remains a crucial concern throughout the development lifecycle of Large Language Models (LLMs), researchers and industrial practitioners have increasingly focused on safeguarding and aligning LLM behaviors with human preferences and ethical standards. LLMs, trained on extensive multilingua…