Teaching LLMs to Refine with Tools
This addresses the limitation of non-correcting behaviors in LLM refinement for AI researchers, though it appears incremental as it builds on existing refinement and tool-use concepts.
The paper tackles the problem of LLMs refining their responses by introducing CaP, a method that uses external tools to refine chain-of-thought responses, achieving effective cross-reasoning refinement and efficient inference.
Large language models (LLMs) can refine their responses based on feedback, enabling self-improvement through iterative training or test-time refinement. However, existing methods predominantly focus on refinement within the same reasoning format, which may lead to non-correcting behaviors. We propose CaP, a novel approach that uses external tools to refine chain-of-thought (CoT) responses generated by the same or other LLMs. CaP employs a two-stage training process: supervised fine-tuning followed by preference optimization with DPO variants. Our observations highlight the critical role of preference optimization in enabling effective refinement. Additionally, we compare several sampling strategies to leverage CoT and tools at inference time. Experimental results demonstrate CaP's potential for effective cross-reasoning refinement and efficient inference.