基于RAG的大模型智能体技术在电影行业的应用探索  

Exploration on the application of RAG⁃based large model agent in the film industry

作　　者：曾真 ZENG Zhen(China Research Institute of Film Science&Technology(Test Institute of Film Technical Quality))

机构地区：[1]中国电影科学技术研究所(中央宣传部电影技术质量检测所),北京100086

出　　处：《现代电影技术》2025年第3期38-44,共7页Advanced Motion Picture Technology

基　　金：中国电影科学技术研究所(中央宣传部电影技术质量检测所)基本科研业务费项目“基于《现代电影技术》期刊内容质量及编审能力提升研究”(2024-DKS-11)。

摘　　要：为探索解决当前创作者使用第三方大模型工具无法充分利用自身数字资源,难以实现电影垂直领域行业定制问题,本文分析了电影行业的人工智能(AI)技术行业定制需求及智能体结合大模型面临的技术挑战,提出一个基于本地开源大模型和检索增强生成(RAG)的智能体(AI Agent)技术框架,并利用本地deepseek-r1:14b创建了一个剧本助手。剧本助手、本地开源大模型、DeepSeek网页版的对比实验测试结果显示,对大模型智能体的设置及RAG的应用,可提升大模型智能体的任务表现。研究表明,基于RAG的大模型智能体能够实现创作者自有数字资源的复用,可为电影行业提供新工具和新方法。To address the current challenges faced by creators who are unable to fully utilize their own digital resources when using third-party large model tools,and to overcome the difficulties in achieving industry-specific customization in the film sector,this paper analyzes the customization requirements of the artificial intelligence(AI)technology in the film indus-try and the technical challenges encountered by LM Agent.This paper proposes a technical framework for LM Agent based on local open source large models and Retrieval-Augmented Generation(RAG).Utilizing the local deepseek-r1:14b model,we developed a script assistant.The comparative experimental tests results of the script assistant,the local open source large model,and the DeepSeek web version demonstrate that the configuration of the LM Agent and the application of RAG can enhance the task performance of the LM Agent.The research indicates that RAG-based LM Agent can realize the reuse of creators'own digital resources,offering new tools and methods for the film industry.

关 键 词：智能体 大模型 RAG 多模态 电影 

分 类 号：TP39[自动化与计算机技术—计算机应用技术] J91[自动化与计算机技术—计算机科学与技术]