LLM-as-an-Investigator: Evidence-First Reasoning for Robust Interactive Problem Diagnosis

Fabrizio Marozzo, Pietro Liò

Large language models (LLMs) are increasingly used as interactive assistants for technical problem solving. However, when users provide incomplete descriptions or plausible but unv…