基于大型语言模型的异构多智能体系统

打开文本图片集

Heterogeneous Multi-agent System Based on Large Language Models

WANG Hang HUANG Xiaoping GAN Guocao RAN Ying

China Travelsky Information Technology Co.， Ltd.， Chongqing Branch， Chongqing 401122， China

Abstract： Aimingat thechallngesfacedbyenterprises duringdigital tansformation whenadoptingasingle-agentframework toachieve intellgentsolutionsacrosultiplebusiesssenarios，thispaperproposesheterogeneousmulti-agentsstembasedon largelanguagemodels.Thesystemcategorizesagentsbasedonnterprisebusiessdomaisandconstructsaherarchicalarchtecture forheterogeneousulti-agentsystems，enablingagentssstemtocolaborateinalayeredmaner.Withinthisarchitecture，amaster agentisrsposiblefortassschssinsdaicogoicegistratioviedstrbtiodsultatio Domain-specificagentsoperateindependentlybasedonamicroservicesarcitecture，enancingtheeficiencyandflexibilityofthe heterogeneous multi-agentsystem.Experimentalresultsdemonstratethatthesystemnotonlyaccomplishesspecifictasksacros numerous and highly diverse scenarios but also autonomouslyselects theagent with thebest execution performance.

Keywords： heterogeneous multi-agent system; arge language models; master agent; domain-specific agents; microservices architecture

0 引言

随着人工智能技术的快速发展，大型语言模型（large language model，LLM）[1]，如 ChatGPT[2]、Deep-Seek[3等，以其卓越的自然语言理解、处理和生成能力，在多个领域展现出广阔的应用前景。（剩余6636字）