远程教育杂志2025，Vol.43Issue(1)：33-45,13.DOI:10.15881/j.cnki.cn33-1304/g4.2025.01.004
大模型驱动的教育多智能体系统应用研究
Research on Educational Multi-Agent Systems Empowered by Large Language Models—Technical Architecture,Current Status,Practical Pathways,and Future Prospects
摘要
Abstract
The rapid advancement of generative large language models(LLMs)has facilitated the widespread application of LLM-powered single-agent systems in educational contexts.However,when addressing educational tasks that require multi-role col-laboration—such as the development of educational resources and simulated collaborative learning in complex environments—single-a-gent systems face significant limitations,including narrow cognitive perspectives and fixed role definitions.These limitations impede their ability to meet the evolving demands of education.This study systematically examines the current applications and future prospects of LLM-powered educational multi-agent systems.First,it establishes a multi-layered technical architecture for educational multi-agent systems,providing a theoretical foundation for related research.Second,the study conducts an in-depth analysis of the current state and typical case studies of educational multi-agent systems across four dimensions:teaching assistance,learning support,educational assessment,and educational research.Third,it systematically explores the internal mechanisms through which multi-agent systems empower education,addressing three key aspects:application mechanisms,technical implementation,and educational effec-tiveness.The study proposes a closed-loop practical framework that encompasses"mechanism design—system modeling—effectiveness evaluation."Finally,the research discusses the future development trends,challenges,and strategies for addressing these challenges in the context of educational multi-agent systems.The findings offer valuable theoretical insights and practical guidance for promoting the deep integration of multi-agent technology with education and advancing innovative developments in intelligent education.关键词
大模型/多智能体系统/生成式人工智能/教育应用/教育变革/智能教育Key words
Large language models/Multi-agent systems/Generative artificial intelligence/Educational transformation/Intelligent education分类
社会科学引用本文复制引用
刘石奇,刘智,段会敏,粟柱,彭晛..大模型驱动的教育多智能体系统应用研究[J].远程教育杂志,2025,43(1):33-45,13.基金项目
本文系国家自然科学基金重点项目"面向智慧教育的多模态模型构建方法"(项目编号:62437002)、国家自然科学基金面上项目"融合情绪感知与归因推理的异步讨论多策略组合干预方法研究"(项目编号:62377016)的研究成果. （项目编号:62437002）
