审计大模型系统答案确定性与灵活性优化研究  

Research on Optimization of Certainty and Flexibility of Answers in Large Model Audit System

作　　者：李鹏伟 周立云 黄佳佳[1,2] 徐超 

机构地区：[1]南京审计大学 [2]中国石油化工集团公司审计部

出　　处：《中国注册会计师》2025年第4期45-51,5,共8页The Chinese Certified Public Accountant

基　　金：国家自然科学基金面上项目“面向国家审计的大模型训练与知识增强研究”(62472227);江苏省高校自然科学研究面上项目“面向审计效能提升的领域知识图谱构建关键技术研究”(23KJB520015)。

摘　　要：近年来,以大语言模型为代表的AIGC技术迅猛发展为审计行业带来了重大机遇,已经成为审计辅助工具和智能问答系统的首选模型。尽管如此,确保大模型系统在审计领域的智能化、专业性和精确度方面仍需进一步深入研究。本文旨在探讨在审计实践中运用大模型时,如何优化系统以保证系统兼容确定性和灵活性答案的输出需求,从而使得系统达到准确、可信、好用,为大模型在审计领域的高效、高质量应用提供了理论支撑和实践指导。In recent years,the rapid development of AIGC technology,represented by Large Language Models,has brought significant opportunities to the auditing profession and has become the preferred model for audit assistance tools and intelligent question-answering systems.Despite this,further indepth research is needed to ensure that large model systems are intelligent,professional,and accurate in the field of auditing.This paper aims to explore how to optimize the system when using large models in auditing practice,ensuring it meets the requirements for deterministic and flexible answers,thereby achieving accuracy,reliability,and usability.This paper provides theoretical support and practical guidance for the efficient and high-quality application of large models in the auditing field.

关 键 词：人工智能 灵活性 确定性 大模型问答系统 RAG 

分 类 号：F239.4[经济管理—会计学]