大模型在科研管理场景下的情报任务测试  

Intelligence Task Testing of Large Models in Scientific Research Management Scenarios

作　　者：孟芾羽 史筱飞 李潇 李雨轩 金馨 MENG Fuyu;SHI Xiaofei;LI Xiao

机构地区：[1]山东省科学技术情报研究院,济南250000

出　　处：《科技创新与应用》2025年第9期21-25,共5页Technology Innovation and Application

基　　金：山东省发改委“十五五”规划前期研究重大课题(kj04)。

摘　　要：大模型的发展已赋能百行千业,为各行业带来机遇的同时,也对政府的治理效能、工作路径提出新的挑战和要求,科研管理场景下的大模型应用亟待探索。根据实际的科研管理工作,拆解出情报信息感知、情报信息搜集、情报文本挖掘、情报信息组织和发展趋势预测典型任务,对国内主流模型进行任务测试,分析模型在科研管理场景下情报的任务完成情况,为在工作中运用模型提出参考。The development of LLMs empowers numerous industries,presenting opportunities to all sectors.Simultaneously,it poses new challenges and demands on the governance efficiency and working approaches of the government.The application of LLMs in the management context awaits urgent exploration.Based on the actual work,typical tasks such as intelligence information perception,intelligence information collection,intelligence text mining,intelligence information organization,and trend prediction are decomposed.Task tests are conducted on domestic mainstream LLMs,and the completion of intelligence tasks by the models in the research management scenario is analyzed to offer references for the application of models in work.

关 键 词：大语言模型 生成式AI 科研管理 应用测试 情报信息 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]