Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards
This addresses the scalability issue in enhancing reasoning capabilities for language model developers, though it is incremental as it builds on existing fine-tuning methods.
The paper tackles the problem of improving mathematical reasoning in large language models without costly human-authored rationales by proposing Self-Explore, which uses fine-grained rewards from identifying the first wrong step in rationales, resulting in average improvements of 11.57% on GSM8K and 2.89% on MATH compared to supervised fine-tuning.
Training on large amounts of rationales (i.e., CoT Fine-tuning) is effective at improving the reasoning capabilities of large language models (LLMs). However, acquiring human-authored rationales or augmenting rationales from proprietary models is costly and not scalable. In this paper, we study the problem of whether LLMs could self-improve their reasoning capabilities. To this end, we propose Self-Explore, where the LLM is tasked to explore the first wrong step (i.e., the first pit) within the rationale and use such signals as fine-grained rewards for further improvement. On the GSM8K and MATH test set, Self-Explore achieves 11.57% and 2.89% improvement on average across three LLMs compared to supervised fine-tuning (SFT). Our code is available at https://github.com/hbin0701/Self-Explore.