Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AgenticRAG: Tool-Augmented Foundation Models for Zero-Shot Explainable Recommender Systems

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

TL;DR

AgenticRAG presents a novel framework that couples retrieval-augmented generation with external tool invocation and chain-of-thought reasoning to deliver zero-shot, explainable recommendations. By maintaining user/item memories, leveraging a multi-modal knowledge base, and dynamically invoking tools, the approach achieves consistent improvements over baselines on three real-world datasets while offering transparent, step-by-step rationales. Key contributions include a tool-augmented agent architecture, effective application of chain-of-thought prompting in recommendation tasks, and comprehensive empirical validation with ablations and interpretability studies. The work has practical impact by enabling scalable, explainable recommendations in dynamic environments where catalogs evolve and real-time data matters.

Abstract

Foundation models have revolutionized artificial intelligence, yet their application in recommender systems remains limited by reasoning opacity and knowledge constraints. This paper introduces AgenticRAG, a novel framework that combines tool-augmented foundation models with retrieval-augmented generation for zero-shot explainable recommendations. Our approach integrates external tool invocation, knowledge retrieval, and chain-of-thought reasoning to create autonomous recommendation agents capable of transparent decision-making without task-specific training. Experimental results on three real-world datasets demonstrate that AgenticRAG achieves consistent improvements over state-of-the-art baselines, with NDCG@10 improvements of 0.4\% on Amazon Electronics, 0.8\% on MovieLens-1M, and 1.6\% on Yelp datasets. The framework exhibits superior explainability while maintaining computational efficiency comparable to traditional methods.

AgenticRAG: Tool-Augmented Foundation Models for Zero-Shot Explainable Recommender Systems

TL;DR

AgenticRAG presents a novel framework that couples retrieval-augmented generation with external tool invocation and chain-of-thought reasoning to deliver zero-shot, explainable recommendations. By maintaining user/item memories, leveraging a multi-modal knowledge base, and dynamically invoking tools, the approach achieves consistent improvements over baselines on three real-world datasets while offering transparent, step-by-step rationales. Key contributions include a tool-augmented agent architecture, effective application of chain-of-thought prompting in recommendation tasks, and comprehensive empirical validation with ablations and interpretability studies. The work has practical impact by enabling scalable, explainable recommendations in dynamic environments where catalogs evolve and real-time data matters.

Abstract

Foundation models have revolutionized artificial intelligence, yet their application in recommender systems remains limited by reasoning opacity and knowledge constraints. This paper introduces AgenticRAG, a novel framework that combines tool-augmented foundation models with retrieval-augmented generation for zero-shot explainable recommendations. Our approach integrates external tool invocation, knowledge retrieval, and chain-of-thought reasoning to create autonomous recommendation agents capable of transparent decision-making without task-specific training. Experimental results on three real-world datasets demonstrate that AgenticRAG achieves consistent improvements over state-of-the-art baselines, with NDCG@10 improvements of 0.4\% on Amazon Electronics, 0.8\% on MovieLens-1M, and 1.6\% on Yelp datasets. The framework exhibits superior explainability while maintaining computational efficiency comparable to traditional methods.

Paper Structure

This paper contains 27 sections, 20 equations, 4 figures, 4 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Problem Formulation and Motivation
  3. Problem Definition
  4. Motivating Examples
  5. Related Work
  6. Agent-Based Recommender Systems
  7. Retrieval-Augmented Generation in Recommendations
  8. Tool-Enhanced Language Models
  9. Methodology
  10. Framework Overview
  11. RAG-Enhanced Agent Architecture
  12. Tool Invocation System
  13. Chain-of-Thought Reasoning
  14. Agent Collaboration Mechanism
  15. Experimental Setup
  16. ...and 12 more sections

Figures (4)

  • Figure 1: Overall architecture of the AgenticRAG framework showing the integration of RAG, tool invocation, and chain-of-thought reasoning components.
  • Figure 2: Tool usage frequency across different datasets showing that similarity computation and sentiment analysis are the most frequently invoked tools.
  • Figure 3: User evaluation of explanation quality across three dimensions, showing significant improvements with AgenticRAG.
  • Figure 4: Case study showing the step-by-step reasoning process for a complex user query, demonstrating how AgenticRAG integrates multiple information sources.