审计大模型的构建及应用研究——以员工违规经商办企业专项审计为例
Research on the Construction and Application of an Audit Large Language Model——A case study of a special audit of employees engaging in business or running enterprises against regulations
陈亚盛 1蒋礼蔚 1单敏 2庄平2
作者信息
- 1. 厦门大学管理学院,邮政编码:361000
- 2. 国网福建省电力有限公司,邮政编码:350003
- 折叠
摘要
大语言模型等人工智能技术的飞速发展为提高内部审计质效带来了契机.本文设计基于"提示工程+检索增强生成+超参数调整"的审计大模型构建框架,并以员工违规经商办企业专项审计为例,在贯穿审计全流程的九个审计工作场景中,探索并实现了大语言模型在内部审计中的应用.研究结果表明,本文所设计的构建框架可以帮助审计人员在实务中构建与应用审计大模型,为将大语言模型应用于审计工作提供实现路径的参考.
Abstract
The rapid development of artificial intelligence technologies such as large language models has brought opportunities to improve the quality and efficiency of internal audits.This paper designs a framework for constructing an audit large language model based on"Prompt Engineering+Retrieval-Augmented Generation+Hyperparameter Tuning"and takes the special audit of employees engaging in business or running enterprises against regulations as an example.It explores and realizes the application of large language models in internal audits in nine audit work scenarios throughout the entire audit process.The research results show that the construction framework designed in this paper can help auditors build and apply audit large language models in practice,providing reference for the path to applying large language models to audit work.
关键词
审计/大语言模型/提示工程/检索增强生成/超参数调整Key words
audit/large language model/prompt engineering/retrieval-augmented generation/hyperparameter tuning引用本文复制引用
基金项目
国家自然科学基金面上项目(72172132)
出版年
2024
NETL
NSTL
万方数据