Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AgentRec: Next-Generation LLM-Powered Multi-Agent Collaborative Recommendation with Adaptive Intelligence

Bo Ma, Hang Li, ZeHua Hu, XiaoFan Gui, LuYao Liu, Simon Lau

TL;DR

AgentRec tackles dynamic user preferences, multi-objective optimization, and real-time adaptation in conversational recommendation by deploying a hierarchical, four-agent LLM framework (Conversation Understanding, Preference Modeling, Context Awareness, Dynamic Ranking) with an adaptive coordination module. A three-tier learning strategy enables rapid responses for simple queries, intelligent reasoning for moderate cases, and deep collaboration for complex scenarios. Empirical results on DuRecDial, DuRecDial 2.0, and MultiWOZ show consistent improvements in conversation success rate (+2.8%), NDCG@10 (+1.9%), and conversation efficiency (+3.2%) while keeping computational costs comparable to baselines. The approach demonstrates that adaptive, multi-agent coordination can surpass single-agent LLM-based methods, offering a scalable path toward cross-domain, richer inter-agent communication in next-generation conversational recommender systems.

Abstract

Interactive conversational recommender systems have gained significant attention for their ability to capture user preferences through natural language interactions. However, existing approaches face substantial challenges in handling dynamic user preferences, maintaining conversation coherence, and balancing multiple ranking objectives simultaneously. This paper introduces AgentRec, a next-generation LLM-powered multi-agent collaborative recommendation framework that addresses these limitations through hierarchical agent networks with adaptive intelligence. Our approach employs specialized LLM-powered agents for conversation understanding, preference modeling, context awareness, and dynamic ranking, coordinated through an adaptive weighting mechanism that learns from interaction patterns. We propose a three-tier learning strategy combining rapid response for simple queries, intelligent reasoning for complex preferences, and deep collaboration for challenging scenarios. Extensive experiments on three real-world datasets demonstrate that AgentRec achieves consistent improvements over state-of-the-art baselines, with 2.8\% enhancement in conversation success rate, 1.9\% improvement in recommendation accuracy (NDCG@10), and 3.2\% better conversation efficiency while maintaining comparable computational costs through intelligent agent coordination.

AgentRec: Next-Generation LLM-Powered Multi-Agent Collaborative Recommendation with Adaptive Intelligence

TL;DR

AgentRec tackles dynamic user preferences, multi-objective optimization, and real-time adaptation in conversational recommendation by deploying a hierarchical, four-agent LLM framework (Conversation Understanding, Preference Modeling, Context Awareness, Dynamic Ranking) with an adaptive coordination module. A three-tier learning strategy enables rapid responses for simple queries, intelligent reasoning for moderate cases, and deep collaboration for complex scenarios. Empirical results on DuRecDial, DuRecDial 2.0, and MultiWOZ show consistent improvements in conversation success rate (+2.8%), NDCG@10 (+1.9%), and conversation efficiency (+3.2%) while keeping computational costs comparable to baselines. The approach demonstrates that adaptive, multi-agent coordination can surpass single-agent LLM-based methods, offering a scalable path toward cross-domain, richer inter-agent communication in next-generation conversational recommender systems.

Abstract

Interactive conversational recommender systems have gained significant attention for their ability to capture user preferences through natural language interactions. However, existing approaches face substantial challenges in handling dynamic user preferences, maintaining conversation coherence, and balancing multiple ranking objectives simultaneously. This paper introduces AgentRec, a next-generation LLM-powered multi-agent collaborative recommendation framework that addresses these limitations through hierarchical agent networks with adaptive intelligence. Our approach employs specialized LLM-powered agents for conversation understanding, preference modeling, context awareness, and dynamic ranking, coordinated through an adaptive weighting mechanism that learns from interaction patterns. We propose a three-tier learning strategy combining rapid response for simple queries, intelligent reasoning for complex preferences, and deep collaboration for challenging scenarios. Extensive experiments on three real-world datasets demonstrate that AgentRec achieves consistent improvements over state-of-the-art baselines, with 2.8\% enhancement in conversation success rate, 1.9\% improvement in recommendation accuracy (NDCG@10), and 3.2\% better conversation efficiency while maintaining comparable computational costs through intelligent agent coordination.

Paper Structure

This paper contains 19 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Conversational Recommender Systems
  4. Multi-Agent Recommender Systems
  5. Adaptive Learning in Recommendation
  6. AgentRec Framework
  7. Architecture Overview
  8. Specialized Agent Design
  9. Conversation Understanding Agent
  10. Preference Modeling Agent
  11. Context Awareness Agent
  12. Dynamic Ranking Agent
  13. Adaptive Coordination Mechanism
  14. Three-Tier Learning Strategy
  15. Experimental Evaluation
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Overall architecture of AgentRec framework showing the four specialized LLM-powered agents and adaptive intelligence coordination mechanism. Each agent processes specific aspects of the conversational recommendation task while the coordinator dynamically weights their contributions through hierarchical agent networks.
  • Figure 2: Three-tier learning strategy of AgentRec showing dynamic query routing based on complexity analysis. Different tiers handle queries with varying computational requirements to optimize both response time and recommendation quality through adaptive intelligence.