Value-Rich, Differently Organized: Measuring Value Expression and Organization in AI-Generated Debates.

arXiv:2610.28604v1 Announce Type: new Abstract: Current evaluation of LLM-generated debate largely focuses on argumentative quality, leaving the value structure of generated arguments underexamined. We introduce a diagnostic metric framework for comparing how