ARAG: Agentic Retrieval Augmented Generation for Personalized Recommendation

📄 arXiv: 2506.21931v2 📥 PDF

作者: Reza Yousefi Maragheh, Pratheek Vadla, Priyank Gupta, Kai Zhao, Aysenur Inan, Kehui Yao, Jianpeng Xu, Praveen Kanumala, Jason Cho, Sushant Kumar

分类: cs.IR, cs.AI, cs.CL, cs.MA

发布日期: 2025-06-27 (更新: 2025-08-11)


💡 一句话要点

提出ARAG框架以解决个性化推荐中的用户偏好捕捉问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 个性化推荐 检索增强生成 多智能体系统 用户偏好理解 自然语言推理 上下文总结 推荐排序

📋 核心要点

  1. 现有的RAG方法依赖静态检索,无法动态捕捉用户的细微偏好,导致推荐效果不佳。
  2. ARAG框架通过引入多智能体协作机制,利用四个专门的智能体来理解用户偏好并生成推荐。
  3. 实验结果表明,ARAG在多个数据集上显著优于标准RAG和基于时效性的基线,提升效果显著。

📝 摘要(中文)

检索增强生成(RAG)在推荐系统中展现出潜力,但现有方法多依赖静态检索启发式,未能有效捕捉动态推荐场景中的用户偏好。本文提出ARAG,一个集成多智能体协作机制的个性化推荐框架,利用四个专门的基于大语言模型的智能体,分别负责用户理解、自然语言推理、上下文总结和项目排序。通过在三个数据集上的评估,ARAG在NDCG@5和Hit@5上分别提升了42.1%和35.5%,显示出智能体推理在检索增强推荐中的有效性。

🔬 方法详解

问题定义:本文旨在解决现有RAG方法在个性化推荐中无法动态捕捉用户偏好的问题,导致推荐效果不理想。

核心思路:ARAG框架通过引入多智能体协作机制,利用不同智能体分别处理用户理解、语义推理、上下文总结和推荐排序,从而提升推荐的个性化和准确性。

技术框架:ARAG的整体架构包括四个主要模块:用户理解智能体、自然语言推理智能体、上下文总结智能体和项目排序智能体,形成一个协同工作流程。

关键创新:ARAG的核心创新在于将智能体推理机制整合到检索增强推荐中,显著提升了对用户动态偏好的捕捉能力,与传统静态检索方法形成鲜明对比。

关键设计:在设计中,ARAG使用了特定的损失函数来优化智能体之间的协作,同时在网络结构上采用了基于大语言模型的架构,以确保对用户意图的准确理解和推荐的相关性。

📊 实验亮点

ARAG在三个数据集上的实验结果显示,NDCG@5提升了42.1%,Hit@5提升了35.5%,显著优于标准RAG和基于时效性的基线,验证了智能体推理在个性化推荐中的有效性和优势。

🎯 应用场景

ARAG框架在个性化推荐系统中具有广泛的应用潜力,能够为电商、社交媒体和内容推荐等领域提供更为精准的用户体验。通过动态捕捉用户偏好,ARAG能够提升用户满意度和平台的转化率,具有重要的实际价值和未来影响。

📄 摘要(原文)

Retrieval-Augmented Generation (RAG) has shown promise in enhancing recommendation systems by incorporating external context into large language model prompts. However, existing RAG-based approaches often rely on static retrieval heuristics and fail to capture nuanced user preferences in dynamic recommendation scenarios. In this work, we introduce ARAG, an Agentic Retrieval-Augmented Generation framework for Personalized Recommendation, which integrates a multi-agent collaboration mechanism into the RAG pipeline. To better understand the long-term and session behavior of the user, ARAG leverages four specialized LLM-based agents: a User Understanding Agent that summarizes user preferences from long-term and session contexts, a Natural Language Inference (NLI) Agent that evaluates semantic alignment between candidate items retrieved by RAG and inferred intent, a context summary agent that summarizes the findings of NLI agent, and an Item Ranker Agent that generates a ranked list of recommendations based on contextual fit. We evaluate ARAG accross three datasets. Experimental results demonstrate that ARAG significantly outperforms standard RAG and recency-based baselines, achieving up to 42.1% improvement in NDCG@5 and 35.5% in Hit@5. We also, conduct an ablation study to analyse the effect by different components of ARAG. Our findings highlight the effectiveness of integrating agentic reasoning into retrieval-augmented recommendation and provide new directions for LLM-based personalization.