When Absence Is Evidence: Evaluating Completeness-Sensitive Negative Reasoning in Large Language Models

Byoungjae Min, Kennedy Edemacu, Sae-Hong Cho +3

Large language models (LLMs) are often asked whether something is absent from a record, list, or retrieved context. Yet non-observation licenses a negative answer only when evidenc…