A Survey of Large Language Model Empowered Agents for Recommendation and Search: Towards Next-Generation Information Retrieval

TL;DR

The paper surveys the rapidly growing use of large language model (LLM) agents to power recommendation and search systems, arguing that their reasoning, memory, and tool-using capabilities can address longstanding information retrieval challenges. It introduces a taxonomy that categorizes LLM-agent roles into four RS domains (interaction, representation, integration, environment simulation) and five search domains (task decomposition, query rewriting, action execution, results synthesis, user simulation), and surveys representative works in each area. It also discusses embodied LLM agents that interact with cyber environments, outlining their potential to enable lifelong learning, one-for-all adaptability, and on-device personalization, along with current limitations and future directions. The paper further highlights open problems such as hallucinations, bias, deployment costs, and privacy, and proposes research avenues to advance robust, scalable IR with LLM-enabled agents.

Abstract

Information technology has profoundly altered the way humans interact with information. The vast amount of content created, shared, and disseminated online has made it increasingly difficult to access relevant information. Over the past two decades, recommender systems and search (collectively referred to as information retrieval systems) have evolved significantly to address these challenges. Recent advances in large language models (LLMs) have demonstrated capabilities that surpass human performance in various language-related tasks and exhibit general understanding, reasoning, and decision-making abilities. This paper explores the transformative potential of LLM agents in enhancing recommender and search systems. We discuss the motivations and roles of LLM agents, and establish a classification framework to elaborate on the existing research. We highlight the immense potential of LLM agents in addressing current challenges in recommendation and search, providing insights into future research directions. This paper is the first to systematically review and classify the research on LLM agents in these domains, offering a novel perspective on leveraging this advanced AI technology for information retrieval. To help understand the existing works, we list the existing papers on LLM agent based recommendation and search at this link: https://github.com/tsinghua-fib-lab/LLM-Agent-for-Recommendation-and-Search.

Figures (4)

• Figure 1: Schematic diagram of the three core modules of agent
• Figure 2: Illustration of existing work of LLM agents for recommendation and search.
• Figure 3: Four domains of LLM agent's role in recommendation tasks
• Figure 4: Five domains of LLM agent's role in search tasks