[3Xin2] Poster session 1

[3Xin2-108] How Pre-Trained Models Benefit Decision-Making in Open-World Games

〇Shiyao Ding¹, Takayuki Ito¹ (1.Kyoto University)

Keywords:Multi-Agent

The purpose of this study is to explore how pre-trained models, such as large language models, support and improve agent decision-making in open-world games. Open-world games require agents to make numerous choices and strategic decisions due to the high degree of environmental freedom. This study discusses how several representative pre-trained models can be used to support agents' decision-making in games and evaluates their effectiveness.