Saved in:
Bibliographic Details
Main Authors: Xu, Xiaojie, Li, Zongyuan, Lu, Chang, Qi, Runnan, Ni, Yanan, Jiang, Lumin, Liu, Xiangbei, Zhang, Xuebo, Fang, Yongchun, Huang, Kuihua, Guo, Xian, Wu, Zhanghua, Li, Zhenya
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.13388
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914463643336704
author Xu, Xiaojie
Li, Zongyuan
Lu, Chang
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Liu, Xiangbei
Zhang, Xuebo
Fang, Yongchun
Huang, Kuihua
Guo, Xian
Wu, Zhanghua
Li, Zhenya
author_facet Xu, Xiaojie
Li, Zongyuan
Lu, Chang
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Liu, Xiangbei
Zhang, Xuebo
Fang, Yongchun
Huang, Kuihua
Guo, Xian
Wu, Zhanghua
Li, Zhenya
contents StarCraft II is a complex and dynamic real-time strategy (RTS) game environment, which is very suitable for artificial intelligence and reinforcement learning research. To address the problem of Large Language Model(LLM) learning in complex environments through self-reflection, we propose a Reflection of Episodes(ROE) framework based on expert experience and self-experience. This framework first obtains key information in the game through a keyframe selection method, then makes decisions based on expert experience and self-experience. After a game is completed, it reflects on the previous experience to obtain new self-experience. Finally, in the experiment, our method beat the robot under the Very Hard difficulty in TextStarCraft II. We analyze the data of the LLM in the process of the game in detail, verified its effectiveness.
format Preprint
id arxiv_https___arxiv_org_abs_2502_13388
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reflection of Episodes: Learning to Play Game from Expert and Self Experiences
Xu, Xiaojie
Li, Zongyuan
Lu, Chang
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Liu, Xiangbei
Zhang, Xuebo
Fang, Yongchun
Huang, Kuihua
Guo, Xian
Wu, Zhanghua
Li, Zhenya
Artificial Intelligence
StarCraft II is a complex and dynamic real-time strategy (RTS) game environment, which is very suitable for artificial intelligence and reinforcement learning research. To address the problem of Large Language Model(LLM) learning in complex environments through self-reflection, we propose a Reflection of Episodes(ROE) framework based on expert experience and self-experience. This framework first obtains key information in the game through a keyframe selection method, then makes decisions based on expert experience and self-experience. After a game is completed, it reflects on the previous experience to obtain new self-experience. Finally, in the experiment, our method beat the robot under the Very Hard difficulty in TextStarCraft II. We analyze the data of the LLM in the process of the game in detail, verified its effectiveness.
title Reflection of Episodes: Learning to Play Game from Expert and Self Experiences
topic Artificial Intelligence
url https://arxiv.org/abs/2502.13388