On the Credibility of Evaluating LLMs using Survey Questions

Jindřich Libovický

Recent studies evaluate the value orientation of large language models (LLMs) using adapted social surveys, typically by prompting models with survey questions and comparing their…