LLM-Agent-for-Recommendation-and-Search

An index for papers on large language model agents for recommendation and search.

Please find more details in our survey paper: A Survey of Large Language Model Empowered Agents for Recommendation and Search: Towards Next-Generation Information Retrieval.

Please cite our survey paper if you find this index helpful.

@article{zhang2025survey,
  title={A Survey of Large Language Model Empowered Agents for Recommendation and Search: Towards Next-Generation Information Retrieval},
  author={Zhang, Yu and Qiao, Shutong and Zhang, Jiaqi and Lin, Tzu-Heng and Gao, Chen and Li, Yong},
  journal={arXiv preprint arXiv:2503.05659},
  year={2025}
}

Recommendation

Domain	Paper	What agents can do (ability)
Interaction	RAH! RecSys--Assistant--Human: A Human-Centered Recommendation Framework With LLM Agents [paper]	Assists users in customized recommendations and feedback (learning–act–critic loop with multi-agent coordination)
Interaction	Let Me Do It For You: Towards LLM Empowered Recommendation via Tool Learning (ToolRec) [paper]	Uses tools for specific recommendation tasks (tool-calling + execution planning)
Interaction	RecAI: Leveraging Large Language Models for Next-Generation Recommender Systems [paper]	Uses LLM as an interface/orchestrator for traditional recommendation tools (plan → call tools → compose response)
Interaction	Automated Interactive Domain-Specific Conversational Agents that Understand Human Dialogs (AutoConcierge) [paper]	Conducts real conversations with users (dialogue understanding + preference elicitation + KB querying)
Interaction	FLOW: A Feedback LOop FrameWork for Simultaneously Enhancing Recommendation and User Agents [paper]	Builds a feedback loop between recommender and user agents (iterative refinement via simulated feedback)
Interaction	iAgent: LLM Agent as a Shield between User and Recommender Systems [paper]	Acts as a shielding agent that mediates user–RS interaction by rewriting, filtering, and sanitizing user inputs to protect privacy without modifying the underlying recommender
Interaction	AgentSociety Challenge: Designing LLM Agents for User Modeling and Recommendation on Web Platforms [paper]	Introduces a multi-agent challenge and evaluation platform for studying how LLM agents model users and coordinate recommendation behaviors on web environments
Interaction	Thought-Augmented Planning for LLM-Powered Interactive Recommender Agent [paper]	Incorporates explicit thought-augmented reasoning to guide planning and action selection in interactive recommendation dialogues
Interaction	Llm-based conversational recommendation agents with collaborative verbalized experience [paper]	Proposes collaborative verbalized experience as a shared, language-level memory to enhance preference elicitation and response generation in conversational recommendation
Representation	AgentCF: Collaborative learning with autonomous language agents for recommender systems [paper]	Models users and items as autonomous language agents and enables collaborative representation learning through agent-level interaction
Representation	Prospect Personalized Recommendation on Large Language Model-based Agent Platform (Rec4Agentverse) [paper]	Introduces an agent-based platform where items and recommenders are modeled as interacting agents for personalized recommendation
Representation	Knowledge Graph Enhanced Language Agents for Recommendation [paper]	Enhances user representation by integrating knowledge-graph reasoning into language-agent-based modeling for recommendation
Representation	PersonaX: A Recommendation Agent Oriented User Modeling Framework for Long Behavior Sequence [paper]	Constructs explicit persona-oriented user representations from long behavior sequences to support stable and interpretable recommendation
Representation	MR.Rec: Synergizing Memory and Reasoning for Personalized Recommendation Assistant with LLMs [paper]	Improves user representation via long-term memory and reasoning-aware retrieval for personalized recommendation
System	Recmind: Large language model powered agent for recommendation (RecMind) [paper]	Self-inspiring decision-making agent that reuses historical reasoning paths to improve planning
System	Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations (InteRec) [paper]	Integrates an LLM agent as the interaction and decision layer that invokes and coordinates traditional recommender modules
System	Multi-Agent Collaboration Framework for Recommender Systems (MACRec) [paper]	Introduces a multi-agent framework where specialized agents collaboratively perform recommendation tasks under a central coordinator
System	Enhancing Long-Term Recommendation with Bi-level Learnable Large Language Model Planning (BiLLP) [paper]	Improves long-term user retention via LLM-planned reinforcement learning with a high-level planner for long-horizon objectives
System	A Multi-Agent Conversational Recommender System (MACRS) [paper]	Employs multiple conversational agents to control dialogue flow and integrate user feedback for recommendation
System	Personalized Recommendation Systems using Multimodal, Autonomous, Multi-Agent Systems (PAS) [paper]	Utilizes multimodal perception and autonomous multi-agent collaboration to support personalized recommendation
System	Lending Interaction Wings to Recommender Systems with Conversational Agents (CORE) [paper]	Augments recommender systems with conversational agents to enhance interaction and user engagement
System	A Hybrid Multi-Agent Conversational Recommender System with LLM and Search Engine in E-commerce (Hybrid-MACRS) [paper]	Combines LLM agents and search engines to coordinate dialogue, retrieval, and re-ranking in conversational recommendation
System	Tunable LLM-based Proactive Recommendation Agent (TunableAgent) [paper]	Proposes a tunable LLM-based agent that adjusts the level of proactive recommendation according to user preferences and context
System	ARAG: Agentic Retrieval Augmented Generation for Personalized Recommendation [paper]	Introduces an agentic RAG framework where an LLM agent controls retrieval and generation for personalized recommendation
System	Toward Safe and Human-Aligned Game Conversational Recommendation via Multi-Agent Decomposition (MATCHA) [paper]	Decomposes conversational recommendation into multiple collaborating agents to improve safety, alignment, and control
System	Cooperative and Competitive LLM-Based Multi-Agent Systems for Recommendation (CoopCompAgents) [paper]	Studies cooperative and competitive interactions among multiple LLM agents for recommendation decision making
System	Answering Narrative-Driven Recommendation Queries via a Retrieve–Rank Paradigm and the OCG-Agent [paper]	Presents an agentic retrieve–rank pipeline where an LLM agent controls retrieval and ranking for narrative-driven recommendation
System	Multi-Agent Collaborative Filtering: Orchestrating Users and Items for Agentic Recommendations (MACF) [paper]	Proposes multi-agent collaborative filtering with a central orchestrator coordinating user and item agents
System	AgentDR: Dynamic Recommendation with Implicit Item–Item Relations via LLM-based Agents [paper]	Introduces a tool-integrated LLM agent for relational reasoning over implicit item–item relations with scalable ranking
Simulation	On Generative Agents in Recommendation (Agent4Rec) [paper]	Uses generative LLM-based agents to simulate user interactions for training and evaluating recommender systems
Simulation	RecAgent: A Novel Simulation Paradigm for Recommender Systems [paper]	Proposes an agent-based user simulator with profile, memory, and action modules to simulate user behaviors in recommender systems
Simulation	SUBER: An RL Environment with Simulated Human Behavior for Recommender Systems [paper]	Constructs an RL environment with simulated human feedback, where language models approximate user responses rather than act as autonomous agents
Simulation	KuaiSim: A Comprehensive Simulator for Recommender Systems [paper]	Provides a reinforcement-learning-based simulation environment for training and evaluating recommender systems
Simulation	CheatAgent: Attacking LLM-Empowered Recommender Systems via LLM Agent [paper]	Demonstrates how adversarial LLM-based user agents can manipulate and attack LLM-empowered recommender systems
Simulation	Agentic Feedback Loop Modeling Improves Recommendation and User Simulation (AFL) [paper]	Models an agentic feedback loop that iteratively refines recommender policies and user simulation through mutual feedback
Simulation	AgentRecBench: Benchmarking LLM Agent-based Personalized Recommender Systems [paper]	Introduces a benchmark and evaluation protocol for assessing LLM agent-based personalized recommender systems
Simulation	LLM-Powered User Simulator for Recommender System (LLMUserSim) [paper]	Uses large language models as user simulators to generate realistic interaction trajectories for recommender system training
Simulation	SimUSER: Simulating User Behavior with Large Language Models for Recommender System Evaluation [paper]	Proposes a two-phase LLM-based user simulation framework separating preference elicitation and response generation
Simulation	AgentCF++: Memory-enhanced LLM-based Agents for Popularity-aware Cross-domain Recommendations [paper]	Introduces memory-enhanced LLM-based user agents to model cross-domain user behaviors for popularity-aware recommendation

Search

Role of agent	Paper	What agents can do (ability)
Decomposer	Knowagent: Knowledge-augmented planning for llm-based agents [paper]	Integrates knowledge base for task decomposition and logical action execution
Decomposer	On the Multi-turn Instruction Following for Conversational Web Agents [paper]	Utilizes self-reflection memory enhancement planning for web navigation tasks
Decomposer	Mindsearch: Mimicking human minds elicits deep ai searcher [paper]	Learns from experience to decompose tasks and execute multi-step web search
Decomposer	Autoact: Automatic agent learning from scratch for qa via self-planning [paper]	Auto-learns QA agents via self-instruct and trajectory synthesis with multiple sub-agents
Decomposer	ManuSearch: Democratizing Deep Search in Large Language Models with a Transparent and Open Multi-Agent Framework [paper]	Modular multi-agent framework for deep web reasoning tasks
Decomposer	StepSearch: Igniting LLMs Search Ability via Step-Wise Proximal Policy Optimization [paper]	Step-wise RL optimization for multi-hop QA with dual rewards
Decomposer	Kwaiagents: Generalized information-seeking agent system with large language models [paper]	Hybrid memory bank + toolset for time-sensitive and long-tail information needs
Decomposer	Webdancer: Towards autonomous information seeking agency [paper]	ReAct-style web agent with autonomous multi-step reasoning--action and tool-based web interaction.
Decomposer	Search-o1: Agentic Search-Enhanced Large Reasoning Models [paper]	Agentic search-enhanced reasoning framework that dynamically decomposes reasoning into search queries and retrieves external evidence during inference.
Rewriter	CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models [paper]	A collaborative agent for query optimization, web search, and cited answer generation.
Rewriter	TCAF: a multi-agent approach of thought chain for retrieval augmented generation [paper]	A multi-agent RAG framework with reference constraints for complex question answering.
Rewriter	Doing Personal LAPS: LLM-Augmented Dialogue Construction for Personalized Multi-Session Conversational Search [paper]	Uses LLM-guided self-dialogue to collect personalized data with a preference memory.
Rewriter	Trec ikat 2023: The interactive knowledge assistance track overview [paper]	A PTKB-based personalized conversational search pipeline.
Rewriter	Agent4Ranking: Semantic Robust Ranking via Personalized Query Rewriting Using Multi-Agent LLMs [paper]	Employs multi-role agents for effective query rewriting and robust ranking.
Executor	Llm agents improve semantic code search [paper]	An RAG-based agent ensemble for semantic code search.
Executor	Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools [paper]	A multi-agent deep research system integrating mind-map, web-search, and coding agents.
Executor	PaSa: An LLM Agent for Comprehensive Academic Paper Search [paper]	A dual-agent academic search framework optimized with reinforcement learning.
Synthesizer	PersonaRAG: Enhancing Retrieval-Augmented Generation Systems with User-Centric Agents [paper]	User-centric agents for personalized and generalizable Retrieval-Augmented Generation (RAG).
Synthesizer	ChatCite: LLM agent with human workflow guidance for comparative literature summary [paper]	Uses reflective generation for comparative literature summaries with structured evaluation.
Synthesizer	Agent-G: An Agentic Framework for Graph Retrieval Augmented Generation [paper]	Features a retriever bank and a critic, enabling self-reflection on retrieval-augmented generation.
Synthesizer	Weknow-rag: An adaptive approach for retrieval-augmented generation integrating web search and knowledge graphs [paper]	Integrates web search and knowledge graphs with multi-stage retrieval and self-assessment.
Simulator	Usimagent: Large language models for simulating search users [paper]	An LLM-based simulator that models user search behaviors for IR evaluation.
Simulator	BASES: Large-scale Web Search User Simulation with Large Language Model based Agents [paper]	Uses user-profiled LLM agents to simulate web search users.
Simulator	ChatShop: Interactive Information Seeking with Language Agents [paper]	Conversational search simulators for evaluation with utterance-level analysis.
Simulator	LEMSS: LLM-Based Platform for Multi-Agent Competitive Search Simulation [paper]	A multi-agent platform to simulate competitive search environments and analyze dynamics.