Hierarchical Expert Prompt for Large-Language-Model: An Approach Defeat Elite AI in TextStarCraft II for the First Time

• DOI: 10.48550/arxiv.2502.11122
• Publication Date: 2025-02-16
• Authors: Zongyuan Li; Chang‐Tien Lu; Xiaojie Xu; R. Qi; Yanan Ni; Ling Jiang; Xiangbei Liu; Xuebo Zhang; Yongchun Fang; Kuihua Huang; Guo Xian
• Keywords: Elite

Since the emergence of Large Language Model (LLM), LLM has been widely used in fields such as writing, translating, and searching. However, there is still great potential for LLM-based methods handling complex tasks decision-making StarCraft II environment. To address problems lack relevant knowledge poor control over subtasks varying importance, we propose a Hierarchical Expert Prompt (HEP) LLM. Our method improves understanding game situations through expert-level tactical knowledge, improving processing quality importance hierarchical framework. approach defeated highest level (Elite) standard built-in agent TextStarCraft first time consistently outperformed baseline other difficulties. experiments suggest that proposed practical solution tackling challenges. The replay video can be viewed on https://www.bilibili.com/video/BV1uz42187EF https://youtu.be/dO3PshWLV5M, our codes have open-sourced https://github.com/luchang1113/HEP-LLM-play-StarCraftII.