Table-R1: Region-based Reinforcement Learning for Table Understanding (2505.12415v2)

Abstract: Tables present unique challenges for LLMs due to their structured row-column interactions, necessitating specialized approaches for effective comprehension. While LLMs have demonstrated potential in table reasoning through prompting and techniques like chain-of-thought (CoT) and program-of-thought (PoT), optimizing their performance for table question answering remains underexplored. In this paper, we introduce region-based Table-R1, a novel reinforcement learning approach that enhances LLM table understanding by integrating region evidence into reasoning steps. Our method employs Region-Enhanced Supervised Fine-Tuning (RE-SFT) to guide models in identifying relevant table regions before generating answers, incorporating textual, symbolic, and program-based reasoning. Additionally, Table-Aware Group Relative Policy Optimization (TARPO) introduces a mixed reward system to dynamically balance region accuracy and answer correctness, with decaying region rewards and consistency penalties to align reasoning steps. Experiments show that Table-R1 achieves an average performance improvement of 14.36 points across multiple base models on three benchmark datasets, even outperforming baseline models with ten times the parameters, while TARPO reduces response token consumption by 67.5% compared to GRPO, significantly advancing LLM capabilities in efficient tabular reasoning.