Alignment Revisited: Are Large Language Models Consistent in Stated and Revealed Preferences?

Zhuojun Gu, Quan Wang, Shuchu Han

Recent advances in Large Language Models (LLMs) highlight the need to align their behaviors with human values. A critical, yet understudied, issue is the potential divergence betwe…