arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.20510cs.AI
收藏

通过大师蒸馏实现语言模型中的接地国际象棋推理

Grounded Chess Reasoning in Language Models via Master Distillation

Department of Computer Science, University of Toronto(多伦多大学计算机科学系);Queen’s University(皇后大学);Coolwei AI Lab(Coolwei人工智能实验室)

Zhenwei Tang, Qianfeng Wen, Seth Grief-Albert, Yahya Elgabra, Blair Yang, Honghua Dong, Ashton Anderson

AI总结:

提出大师蒸馏框架,将专家系统推理过程蒸馏为自然语言思维链,使4B参数模型C1在国际象棋中达到48.1%准确率,超越所有开源模型和多数前沿闭源系统。

摘要:

语言模型在训练数据稀缺但定制系统表现出色的专业领域中往往缺乏接地推理能力。我们引入了一个通用框架,将专家系统推理蒸馏为自然语言思维链解释,使紧凑模型能够获得领域专业知识并生成忠实、可解释的解释。我们不仅蒸馏最终输出,还捕获完整的推理过程,将不透明的专家计算转化为透明的逐步解释。我们在国际象棋中展示了这一方法,这是一个语言模型仍然表现不佳的典型推理领域。我们的4B参数模型C1从接近零的基线提升到48.1%的准确率,超越了所有开源模型和大多数前沿闭源系统。值得注意的是,C1超越了其蒸馏教师,并且生成的解决方案的token数比基线少两个数量级。与仅预测最佳走法的先前神经国际象棋方法不同,C1生成可解释的解决方案,揭示战略推理。我们的流程结合了监督微调和强化学习,以及主题平衡的数据采样,以实现全面的战术覆盖。大师蒸馏展示了如何将专家级知识注入紧凑模型,用于优化不足的领域,为在LLM缺乏足够基础能力的情况下解锁RLVR提供了一种方法。

英文摘要:

Language models often lack grounded reasoning capabilities in specialized domains where training data is scarce but bespoke systems excel. We introduce a general framework for distilling expert system reasoning into natural language chain-of-thought explanations, enabling compact models to acquire domain expertise and the ability to generate faithful, grounded explanations. Rather than distilling only final outputs, we capture the full reasoning process, transforming opaque expert computations into transparent, step-by-step explanations. We demonstrate this approach in chess, a canonical reasoning domain where language models continue to underperform. Our 4B parameter model, C1, advances from a near-zero baseline to 48.1\% accuracy, outperforming all open-source models and most frontier proprietary systems. Notably, C1 surpasses its distillation teacher and generates solutions in two orders of magnitude fewer tokens than baselines. Unlike prior neural chess approaches that predict only best moves, C1 generates explainable solutions revealing strategic reasoning. Our pipeline combines supervised fine-tuning and reinforcement learning with theme-balanced data sampling for comprehensive tactical coverage. Master Distillation demonstrates how to inject expert-level knowledge into compact models for under-optimized domains, offering a recipe for unlocking RLVR where LLMs lack sufficient base capabilities.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑