Reflect before Act: Proactive Error Correction in Language Models

Qiuhai Zeng, Sarvesh Rajkumar, Di Wang +2

Large Language Models (LLMs) have demonstrated remarkable capabilities in interactive decision-making tasks, but existing methods often struggle with error accumulation and lack ro…