G-Refer: Graph Retrieval-Augmented Large Language Model for Explainable Recommendation

2025-02-18

AI Summary

This work addresses explainable recommendation by tackling two key challenges: (1) difficulty in extracting collaborative filtering (CF) signals from user-item interaction graphs, and (2) modality mismatch between graph-structured CF representations and large language models (LLMs), leading to inaccurate explanation generation. To resolve these, we propose a hybrid graph retrieval mechanism that explicitly models CF signals from both structural and semantic perspectives; a graph translation framework that converts graph topology and relational information into natural language text; and knowledge pruning combined with retrieval-augmented fine-tuning (RAG) to bridge the modality gap between graph embeddings and LLM inputs. Our approach enables end-to-end joint modeling of CF signals and LLMs. Extensive experiments demonstrate significant improvements over state-of-the-art methods in explanation quality, faithfulness, and generation stability. The code and datasets are publicly available.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Explainable recommendation has demonstrated significant advantages in informing users about the logic behind recommendations, thereby increasing system transparency, effectiveness, and trustworthiness. To provide personalized and interpretable explanations, existing works often combine the generation capabilities of large language models (LLMs) with collaborative filtering (CF) information. CF information extracted from the user-item interaction graph captures the user behaviors and preferences, which is crucial for providing informative explanations. However, due to the complexity of graph structure, effectively extracting the CF information from graphs still remains a challenge. Moreover, existing methods often struggle with the integration of extracted CF information with LLMs due to its implicit representation and the modality gap between graph structures and natural language explanations. To address these challenges, we propose G-Refer, a framework using graph retrieval-augmented large language models (LLMs) for explainable recommendation. Specifically, we first employ a hybrid graph retrieval mechanism to retrieve explicit CF signals from both structural and semantic perspectives. The retrieved CF information is explicitly formulated as human-understandable text by the proposed graph translation and accounts for the explanations generated by LLMs. To bridge the modality gap, we introduce knowledge pruning and retrieval-augmented fine-tuning to enhance the ability of LLMs to process and utilize the retrieved CF information to generate explanations. Extensive experiments show that G-Refer achieves superior performance compared with existing methods in both explainability and stability. Codes and data are available at https://github.com/Yuhan1i/G-Refer.

Problem

• Enhance explainable recommendation system transparency
• Integrate collaborative filtering with large language models
• Bridge modality gap between graph and language

Innovation

• Graph retrieval-augmented LLMs
• Hybrid graph retrieval mechanism
• Knowledge pruning and fine-tuning