表-R1：表格推理的推理時擴展

Table-R1: Inference-Time Scaling for Table Reasoning

May 29, 2025

作者: Zheyuan Yang, Lyuhao Chen, Arman Cohan, Yilun Zhao

cs.AI

摘要

在本研究中，我們首次探討了在表格推理任務上的推理時擴展方法。我們開發並評估了兩種訓練後策略來實現推理時擴展：基於前沿模型推理軌跡的知識蒸餾，以及帶有可驗證獎勵的強化學習（RLVR）。對於知識蒸餾，我們引入了一個由DeepSeek-R1生成的大規模推理軌跡數據集，並用它來微調大型語言模型（LLMs），從而得到Table-R1-SFT模型。對於RLVR，我們提出了任務特定的可驗證獎勵函數，並應用GRPO算法來獲得Table-R1-Zero模型。我們在多樣化的表格推理任務上評估了我們的Table-R1系列模型，包括簡答問答、事實驗證和自由形式問答。值得注意的是，Table-R1-Zero模型在僅使用7B參數的LLM的情況下，其性能匹配甚至超越了GPT-4.1和DeepSeek-R1。此外，它還展現出對域外數據集的強大泛化能力。廣泛的消融實驗和定性分析揭示了指令微調、模型架構選擇和跨任務泛化的益處，以及在RL訓練過程中基本表格推理技能的湧現。

English

In this work, we present the first study to explore inference-time scaling on table reasoning tasks. We develop and evaluate two post-training strategies to enable inference-time scaling: distillation from frontier model reasoning traces and reinforcement learning with verifiable rewards (RLVR). For distillation, we introduce a large-scale dataset of reasoning traces generated by DeepSeek-R1, which we use to fine-tune LLMs into the Table-R1-SFT model. For RLVR, we propose task-specific verifiable reward functions and apply the GRPO algorithm to obtain the Table-R1-Zero model. We evaluate our Table-R1-series models across diverse table reasoning tasks, including short-form QA, fact verification, and free-form QA. Notably, the Table-R1-Zero model matches or exceeds the performance of GPT-4.1 and DeepSeek-R1, while using only a 7B-parameter LLM. It also demonstrates strong generalization to out-of-domain datasets. Extensive ablation and qualitative analyses reveal the benefits of instruction tuning, model architecture choices, and cross-task generalization, as well as emergence of essential table reasoning skills during RL training.