大模型提示词工程的进展、综述及展望

Progress,review and prospect of large model prompt word engineering are introduced

自2023年起,大型语言模型(LLM)进入了快速发展期,提示词工程愈发受到重视并成为重要研究课题之一。文章首先介绍了提示词工程的定义分类和主要要素;然后进一步探讨了零样本提示、少样本提示、链式思考提示、思维树提示、自动提示工程、检索增强生成和ReAct框架等提示技术,并讨论了它们在不同场景下的应用和效果;最后预测了提示词工程未来可能的发展趋势。

Since 2023,large language models(LLM)have entered a period of rapid development.At present,prompt word engineering has attracted more and more attention and become one of the important topics.This paper first introduces the definition,classification and main elements of prompt word engineering.Then it further discusses the techniques of zero sample prompt,less sample prompt,chain thinking prompt,thought tree prompt,automatic prompt engineering,search enhancement generation and ReAct framework,and discusses their application and effect in different scenarios.Finally,the paper predicts the possible development trend of prompt word engineering in the future.