MADREC: A Multi-Aspect Driven LLM Agent for Explainable and Adaptive Recommendation

2025-10-15

AI Summary

Existing LLM-based recommendation methods predominantly rely on static prompts, failing to capture the dynamic evolution of user preferences and the complexity of interactive behavior. To address this, we propose MADRec—a Multi-dimensional Adaptive Recommendation agent. First, it unsupervisedly extracts fine-grained aspect-level features from user reviews to construct structured, multi-dimensional user–item profiles. Second, it introduces a Self-Feedback mechanism that dynamically switches reasoning strategies when target items are absent. Third, it employs Re-Ranking to enhance input density and jointly generates interpretable recommendations. MADRec is end-to-end integrated with LLMs and supports direct sequential recommendation. Extensive experiments across multiple domains demonstrate that MADRec significantly outperforms both traditional and LLM-based baselines, achieving state-of-the-art performance in both recommendation accuracy and human-evaluated explanation quality. It offers strong interpretability and environment-aware adaptability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Multiagent Systems: Adversarial Agents

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Recent attempts to integrate large language models (LLMs) into recommender systems have gained momentum, but most remain limited to simple text generation or static prompt-based inference, failing to capture the complexity of user preferences and real-world interactions. This study proposes the Multi-Aspect Driven LLM Agent MADRec, an autonomous LLM-based recommender that constructs user and item profiles by unsupervised extraction of multi-aspect information from reviews and performs direct recommendation, sequential recommendation, and explanation generation. MADRec generates structured profiles via aspect-category-based summarization and applies Re-Ranking to construct high-density inputs. When the ground-truth item is missing from the output, the Self-Feedback mechanism dynamically adjusts the inference criteria. Experiments across multiple domains show that MADRec outperforms traditional and LLM-based baselines in both precision and explainability, with human evaluation further confirming the persuasiveness of the generated explanations.

Problem

• Addresses limitations of static LLM-based recommender systems
• Constructs adaptive user profiles from multi-aspect review information
• Enhances recommendation accuracy and explainability through dynamic mechanisms

Innovation

• Unsupervised multi-aspect profile extraction from reviews
• Self-feedback mechanism dynamically adjusts inference criteria
• Re-ranking constructs high-density inputs for recommendations