期刊文献+

大语言模型时代下的信息检索研究发展趋势 被引量:8

The Development of Information Retrieval in the Era of Large Language Model
原文传递
导出
摘要 以ChatGPT为代表的大语言模型带来了人工智能技术的新一轮发展浪潮,获得了广泛的社会关注。大语言模型通过大规模无标注数据预训练、指令微调、人类对齐等关键技术途径,学习到了丰富的世界知识,具有较好的文本理解与生成能力,能够有效求解各种复杂任务。这一重要技术进展对于信息检索领域的发展带来了新的机遇。本文从大语言模型对于已有信息检索架构的改进以及现有检索技术如何改进大语言模型两个方面进行阐述,针对相关科学问题的可行技术方法进行了梳理与展望,探讨大语言模型时代下的信息检索发展趋势,旨在推动信息检索领域的科研进步。 Large language models(LLMs)exemplified by ChatGPT has sparked a new wave of development in artificial intelligence technology and received widespread social attention.By exploringa series of technical approaches(i.e.,massive unsupervised pre-training,instruction tuning,and human alignment),LLMs can effectively encode world knowledge and possess excellent language understanding and generation ability,making it possible to solve various complex tasks via a unified model.The technical advance of LLMs has brought new opportunities for the development of the information retrieval(IR)field.This article will elaborate it on two major aspects:how to enhance existing IR techniques by LLMs and how existing IR techniques can benefit LLMs.We systematically review and summarize feasible technical approaches to key research issues,and discuss how IR techniques would be shaped by LLMs,which aims to better explore the development path of IR field in the era of LLMs.
作者 赵鑫 窦志成 文继荣 Wayne Xin;Zhao Zhicheng;Dou Ji-Rong;Wen Gaoling(School of Artificial Intelligence,Renmin University of China,Beijing 100872)
出处 《中国科学基金》 CSSCI CSCD 北大核心 2023年第5期786-792,共7页 Bulletin of National Natural Science Foundation of China
关键词 大语言模型 信息检索 生成式检索 large language models information retrieval generative retrieval
  • 相关文献

同被引文献114

引证文献8

二级引证文献6

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部