Teaching LLMs to Refine with Tools

Dian Yu, Yuheng Zhang, Jiahao Xu +5

Large language models (LLMs) can refine their responses based on feedback, enabling self-improvement through iterative training or test-time refinement. However, existing methods p…