A Survey of Large Language Models in Economics and Finance:Model Evolution,Academic Applications,and Future Prospects
In late 2022,ChatGPT rapidly gained hundreds of millions of users worldwide due to its remarkable per-formance,sparking global interest in large language models(LLMs).These models offer significant technical support for text comprehension,enhancing the accuracy of unstructured data processing,and demonstrate vast potential for applica-tions in economics and finance.This study provides a comprehensive review of the evolution of LLMs,their advantages and limitations in text understanding,and their academic applications and challenges.First,we outline the development of language models from statistical models to LLMs across four stages,highlighting key technical advancements.We then emphasize the superior semantic understanding and reasoning capabilities of LLMs compared to traditional text analysis methods,while also identifying potential challenges in their use.LLMs can extract critical information from text and evaluate economic agents'behaviors more accurately and rationally.This study explores specific applications of LLMs in economics and finance,including policy text analysis,macroeconomic forecasting,labor market structure analysis,corporate information disclosure interpretation,market sentiment assessment,and asset price prediction.As LLMs continue to evolve,future applications should focus on balancing effectiveness with cost and expanding to analyze multimodal data,such as images,audio,and video.In conclusion,by outlining the overall development trends and aca-demic applications,this study aims to provide practical guidance and references for the use of LLMs,particularly in ana-lyzing issues within China's economic and financial sectors.
万方数据
