Enhancing Commentary Strategies for Imperfect Information Card Games: A Study of Large Language Models in Guandan Commentary

TL;DR

This paper presents a novel RL and LLM-based framework for generating insightful and personalized commentary in the complex, imperfect-information Chinese card game Guandan, outperforming GPT-4 in evaluations.

Contribution

It introduces a new method combining RL and LLMs with Theory of Mind capabilities for improved game commentary in imperfect information settings.

Findings

Framework surpasses GPT-4 in multiple metrics

Enhances commentary quality with ToM and retrieval modules

Achieves detailed, context-aware commentary in Chinese

Abstract

Recent advancements in large language models (LLMs) have unlocked the potential for generating high-quality game commentary. However, producing insightful and engaging commentary for complex games with incomplete information remains a significant challenge. In this paper, we introduce a novel commentary method that combine Reinforcement Learning (RL) and LLMs, tailored specifically for the Chinese card game \textit{Guandan}. Our system leverages RL to generate intricate card-playing scenarios and employs LLMs to generate corresponding commentary text, effectively emulating the strategic analysis and narrative prowess of professional commentators. The framework comprises a state commentary guide, a Theory of Mind (ToM)-based strategy analyzer, and a style retrieval module, which seamlessly collaborate to deliver detailed and context-relevant game commentary in the Chinese language environment. We empower LLMs with ToM capabilities and refine both retrieval and information filtering mechanisms. This facilitates the generation of personalized commentary content. Our experimental results showcase the substantial enhancement in performance achieved by the proposed commentary framework when applied to open-source LLMs, surpassing the performance of GPT-4 across multiple evaluation metrics.