[论文解读] Large Language Models for Information Retrieval: A Survey

Yutao Zhu, Huaying Yuan+8 more|arXiv (Cornell University)|Aug 14, 2023

Topic Modeling被引用 92

一句话总结

本综述评估大语言模型（LLMs）在信息检索（IR）中的应用，覆盖查询改写、检索器、再排序、阅读器和搜索代理，强调方法、挑战与方向。

ABSTRACT

As a primary means of information acquisition, information retrieval (IR) systems, such as search engines, have integrated themselves into our daily lives. These systems also serve as components of dialogue, question-answering, and recommender systems. The trajectory of IR has evolved dynamically from its origins in term-based methods to its integration with advanced neural models. While the neural models excel at capturing complex contextual signals and semantic nuances, thereby reshaping the IR landscape, they still face challenges such as data scarcity, interpretability, and the generation of contextually plausible yet potentially inaccurate responses. This evolution requires a combination of both traditional methods (such as term-based sparse retrieval methods with rapid response) and modern neural architectures (such as language models with powerful language understanding capacity). Meanwhile, the emergence of large language models (LLMs), typified by ChatGPT and GPT-4, has revolutionized natural language processing due to their remarkable language understanding, generation, generalization, and reasoning abilities. Consequently, recent research has sought to leverage LLMs to improve IR systems. Given the rapid evolution of this research trajectory, it is necessary to consolidate existing methodologies and provide nuanced insights through a comprehensive overview. In this survey, we delve into the confluence of LLMs and IR systems, including crucial aspects such as query rewriters, retrievers, rerankers, and readers. Additionally, we explore promising directions, such as search agents, within this expanding field.

研究动机与目标

• 整合并分析将LLMs应用于IR组件（查询改写、检索器、再排序、阅读器）的最新进展。
• 解释LLMs如何被整合进IR流程，并识别实际挑战，如数据稀缺和幻觉问题。
• 讨论技术（提示、微调、基于语料的增强方法）和评估方面。
• 探索未来方向，包括搜索代理和端到端的LLM驱动IR模型。

提出的方法

• 综述与IR相关的背景和LLM概念。
• 对IR模块（查询改写、检索、再排序、阅读器）及其LLM增强变体进行分类。
• 讨论提示策略、微调和知识蒸馏在IR任务中的应用。
• 分析基于语料增强与仅LLM的方法在查询改写中的差异。
• 概览搜索代理（如WebGPT）作为统一的LLM驱动IR范式。

实验结果

研究问题

• RQ1 当前LLMs如何被整合到IR模块（查询改写、检索、再排序、阅读器），以及其收益与局限？
• RQ2 哪些提示策略、训练方案和语料库整合能够在临时和对话式搜索中实现最佳IR性能？
• RQ3 在IR中部署LLMs时的关键挑战（数据稀缺、幻觉、可解释性）是什么，如何缓解？
• RQ4 搜索代理在自动化IR任务和提升用户体验方面的作用与潜力？
• RQ5 哪些未来方向与待解决的问题应引导LLM在IR研究中的发展？

主要发现

• LLMs用于提升对临时和对话式搜索的查询改写，通过扩展和澄清查询。
• 基于语料的LLM改写方法利用外部文档来提高改写的相关性和真实性。
• 存在三种主要的改写方法：提示、微调和知识蒸馏，其中提示在IR应用中较为突出。
• LLMs 通过不仅限于改写的模块对IR做出贡献，包括检索、再排序和阅读器，使响应更具上下文感知和生成能力。
• 搜索代理建模自动化搜索和结果理解，使端到端的LLM驱动IR工作流成为可能（如WebGPT）。
• 该领域强调在传统IR效率（如BM25）与神经语义能力之间取得平衡，以应对数据稀缺和鲁棒性问题。