How language models frame answers to subjective, culturally situated questions
A post announcing “Not What, But How,” a new EMNLP 2026 paper, argues that most benchmarks overlook framing features such as insider positioning.
TLDR
A post announcing the EMNLP 2026 paper “Not What, But How: A Communicative Audit of LLM Response Framing” says people use large language models for subjective, culturally situated questions. It argues that most benchmarks ask whether an answer is correct while overlooking framing features such as insider positioning.
Combined views
607
1 Source, first seen ago