大语言模型AI智能体的设计方法研究

黄悦欣; 周雨琪; 覃京燕; 杨胡尧; 王建元

doi:10.19798/j.cnki.2096-6946.2025.06.002

PDF(7959 KB)

工业工程设计 ›› 2025, Vol. 7 ›› Issue (6) : 6-15. DOI: 10.19798/j.cnki.2096-6946.2025.06.002

设计理论

大语言模型AI智能体的设计方法研究

黄悦欣, 周雨琪, 覃京燕^*, 杨胡尧, 王建元

作者信息 +

Design Methodology for Large Language Model-driven AI Agents

HUANG Yuexin, ZHOU Yuqi, QIN Jingyan^*, YANG Huyao, WANG Jianyuan

Author information +

文章历史 +

摘要

针对通用大语言模型AI智能体的设计方法缺乏系统构建,智能体的设计流程与内容生成设计机制缺乏设计本体论方法指导等问题,构建一种面向垂直领域的大语言模型AI智能体的设计方法。搭建面向垂直领域的多源数据处理体系,构建高质量的语义向量表示;设计混合检索架构,提升检索的相关性和覆盖率;引入链式思维与多路径推理机制,结合提示优化策略,提高生成回答的准确性、可控性与一致性;最终提升垂域大模型AI智能体感知认知、规划决策、执行作业与交互进化能力。以智慧城市系统为例,构建了集城市动态感知、灾害预警推演、城市智能解读、民生智能问答于一体的大模型AI智能体系统。该研究验证了大模型AI智能体设计方法的可行性,为大模型AI智能体的理论研究与实践应用提供借鉴参考。

Abstract

To address the lack of systematic construction in the design methods of AI agents driven by the general-purpose Large Language Model (LLM), and the absence of design ontology guidance for their design workflows and content generation mechanisms, the work aims to establish a design methodology for LLM-driven AI agents in vertical domains. A multi-source data processing framework for vertical domains is established to construct high-quality semantic vector representations. A hybrid retrieval architecture is designed to enhance retrieval relevance and coverage. By incorporating chain-of-thought and multi-path reasoning mechanisms combined with prompt optimization strategies, the accuracy, controllability, and consistency of generated responses are improved. Ultimately, this approach enhances the perception-cognition, planning-decision making, execution, and interactive evolution capabilities of LLM-driven AI agents in vertical domain. Using a Smart City system as a case study, we constructed a platform integrating urban dynamic perception, disaster warning simulation, intelligent urban analytics, and citizen service Q&A.. This study validates the feasibility of the design methodology, providing theoretical and practical insights for the research and application of LLM-driven AI agents.

导出引用

黄悦欣, 周雨琪, 覃京燕, 杨胡尧, 王建元. 大语言模型AI智能体的设计方法研究[J]. 工业工程设计. 2025, 7(6): 6-15 https://doi.org/10.19798/j.cnki.2096-6946.2025.06.002

HUANG Yuexin, ZHOU Yuqi, QIN Jingyan, YANG Huyao, WANG Jianyuan. Design Methodology for Large Language Model-driven AI Agents[J]. Industrial & Engineering Design. 2025, 7(6): 6-15 https://doi.org/10.19798/j.cnki.2096-6946.2025.06.002

中图分类号： J524 TB472

参考文献

[1] NOY S, ZHANG W.Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence[J]. Science,2023,381(6654):187-192.
[2] KASNECI E, SESSLER K, KÜCHEMANN S, et al. ChatGPT for Good?on Opportunities and Challenges of Large Language Models for Education[J]. Learning and Individual Differences,2023,103:102274.
[3] GAO S H, FANG A D, HUANG Y P, et al.Empowering Biomedical Discovery with AI Agents[J]. Cell, 2024, 187(22):6125-6151.
[4] THIRUNAVUKARASU A J, TING D S J, ELANGOVAN K, et al. Large Language Models in Medicine[J]. Nature Medicine, 2023, 29(8): 1930-1940.
[5] LAI J Q, GAN W S, WU J Y, et al.Large Language Models in Law: A Survey[J]. AI Open, 2024, 5: 181-196.
[6] 李鹏,陈少飞,易楚舒,等. 多智能体风险决策理论与方法研究综述[J]. 计算机学报,2025,48(10):2338-2370.
[7] GURURANGAN S,MARASOVIĆ A,SWAYAMDIPTA S,et al.Don't Stop Pretraining: Adapt Language Models to Domains and Tasks[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics,Online:Association for Computational Linguistics, 2020:8342-8360.
[8] EPSTEIN Z,HERTZMANN A,CREATIVITY I O H,et al. Art and the Science of Generative AI[J]. Science,2023,380(6650):1110-1111.
[9] JO A.The Promise and Peril of Generative AI[J]. Nature,2023,614(1):214-216.
[10] YANG T,MEI Y P,XU L,et al.Application of Question Answering Systems for Intelligent Agriculture Production and Sustainable Management:A Review[J]. Resources,Conservation and Recycling,2024,204:107497.
[11] LIN T H,HUANG Y H,PUTRANTO A.Intelligent Question and Answer System for Building Information Modeling and Artificial Intelligence of Things Based on the Bidirectional Encoder Representations from Transformers Model[J]. Automation in Construction,2022,142:104483.
[12] LEE J,YOON W,KIM S,et al.BioBERT:A Pre-Trained Biomedical Language Representation Model for Biomedical Text Mining[J]. Bioinformatics,2020,36(4):1234-1240.
[13] CHEN X,MAO M,LI S,SHANGGUAN H.Debate-Feedback: A Multi-Agent Framework for Efficient Legal Judgment Prediction[C]//Chiruzzo L,Ritter A,Wang L,eds. Proceedings of the 2025 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies (Volume 2:Short Papers).Albuquerque:Association for Computational Linguistics,2025:462-470.
[14] LIU X Y,WANG G,YANG H,et al.FinGPT:Democratizing Internet-Scale Data for Financial Large Language Models. arXiv preprint arXiv:2307.10485.
[15] GUO D,YANG D,ZHANG H,et al.DeepSeek-R1 Incentivizes Reasoning in Llms Through Reinforcement Learning[J]. Nature,2025,645:633-636.
[16] BAI J,BAI S,CHU Y,et al. Qwen Technical Report[J/OL]. arXiv preprint arXiv:2309.16609,2023.
[17] 陈晓东,朱肖霏. DeepSeek技术突围重构全球人工智能产业竞争格局的六个维度:范式与启示[J]. 南京社会科学,2025(7):41-57.
[18] SONG Z,YAN B,LIU Y,et al.Injecting Domain-Specific Knowledge into Large Language Models:A Comprehensive Survey[C]//Findings of the Association for Computational Linguistics: EMNLP 2025. Suzhou:Association for Computational Linguistics,2025:25297-25311. DOI:10.18653/v1/2025.findings-emnlp.1379.
[19] CHEN Z,LIU K,WANG Q,et al.Agent-FLAN:Designing Data and Methods of Effective Agent Tuning for Large Language Models[C]//Findings of the Association for Computational Linguistics,ACL 2024. Bangkok:Association for Computational Linguistics,2024:9354-9366.
[20] ZHANG Q,CHEN S,BEI Y,et al.A Survey of Graph Retrieval-Augmented Generation for Customized Large Language Models. arXiv preprint arXiv:2501.13958.
[21] 张新长,华淑贞,齐霁,等. 新型智慧城市建设与展望:基于AI的大数据、大模型与大算力[J]. 地球信息科学学报,2024,26(4):779-789
[22] 李帅峥,钱正,董正浩,等. 政务大模型的技术应用路径及发展建议[J]. 信息通信技术与政策,2024,50(6):76-82.