Grounded Chess Reasoning in Language Models via Master Distillation

TL;DR

This paper presents a framework for distilling expert reasoning into language models using step-by-step explanations, significantly improving their performance and interpretability in complex domains like chess.

Contribution

It introduces Master Distillation, a method to transfer expert reasoning into language models via full reasoning process distillation, enhancing domain expertise and explainability.

Findings

C1 model achieves 48.1% accuracy in chess, surpassing open-source and many proprietary models.

C1 generates solutions with two orders of magnitude fewer tokens than baselines.

The approach produces faithful, grounded explanations, improving interpretability in reasoning tasks.

Abstract

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.