基于大语言模型的印刷业智能体架构探索

Exploring the Architecture of Intelligent Agents in the Print Industry Based on Large Language Models

本研究梳理大语言模型发展历程,揭示智能体发展脉络和最新研究进展。针对现行印刷行业生产高度离散化、智能化程度有限等现状,本研究提出一种基于大型印刷知识模型的智能体架构(Large Print Knowledge Model Based Agent,LPKM-Based Agent)。系统阐述了架构的系统组成和架构原理;分析了整合LPKM-Based Agent多系统之间的数据统筹利用的方法,强调构建大型印刷知识模型的四项原则,并进行实验验证。实验结果表明,该架构融合多模态的印刷领域数据输入,能够构建自决策、自执行、自诊断的印刷智能制造体系;基于LPKM-Based Agent架构的印刷业生产制造模式较现行智能印厂生产体系优势明显。

In this study,the landmark development process of language models were reviewed,and the development context and latest research progress of agents with the development process of large language models were revealed.In view of the current situation of high discrete production and limited intelligence in the printing industry,an agent architecture based on large print knowledge model based agent(LPKM-Based Agent)was proposed.The system composition and architecture principle of this architecture were systematically expounded and the data utilization method between multiple systems integrating LPKM-based agent was analyzed.Emphasize the four principles of building a large print knowledge model and design experimental verification.The experiment result showed that the architecture integrates multi-modal data input in the printing field,and can build a printing intelligent manufacturing system with self-decision-making,self-execution and self-diagnosis.The production and manufacturing mode of the printing industry based on the LPKM-based agent architecture has obvious advantages over the current intelligent printing factory production system.