Breaking Information Cocoons: A Hyperbolic Graph-LLM Framework for Exploration and Exploitation in Recommender Systems

2024-11-21

AI Summary

Recommender systems often suffer from “information cocoons” due to overreliance on users’ historical preferences, struggling to balance novelty and relevance while lacking explicit modeling of hierarchical content structures and personalized exploration–exploitation trade-offs. To address these limitations, we propose the Spherical Graph–LLM (SG-LLM) framework: a novel hierarchical-aware graph neural network that performs multimodal (image–text) co-alignment in hyperspherical space, integrating LLM-based semantic encoding, hierarchical collaborative filtering, and a tunable exploration strategy enabling user-controllable novelty–familiarity trade-offs. Extensive experiments demonstrate that SG-LLM improves utility by 5.49% and diversity by 11.39% over strong baselines, significantly mitigating information cocoons. The implementation is publicly available.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Modern recommender systems often create information cocoons, restricting users' exposure to diverse content. A key challenge lies in balancing content exploration and exploitation while allowing users to adjust their recommendation preferences. Intuitively, this balance can be modeled as a tree-structured representation, where depth search facilitates exploitation and breadth search enables exploration. However, existing approaches face two fundamental limitations: Euclidean methods struggle to capture hierarchical structures, while hyperbolic methods, despite their superior hierarchical modeling, lack semantic understanding of user and item profiles and fail to provide a principled mechanism for balancing exploration and exploitation. To address these challenges, we propose HERec, a hyperbolic graph-LLM framework that effectively balances exploration and exploitation in recommender systems. Our framework introduces two key innovations: (1) a hierarchical-aware graph-LLM mechanism that jointly aligns textual descriptions with user-item collaborative information in hyperbolic space, and (2) a hierarchical representation structure that enables user-adjustable exploration-exploitation trade-offs. Extensive experiments demonstrate that HERec consistently outperforms both Euclidean and hyperbolic baselines, achieving up to 5.49% improvement in utility metrics and 11.39% increase in diversity metrics, effectively mitigating information cocoons. We open-source our model implementation at https://github.com/Martin-qyma/HERec.

Problem

• Recommendation Systems
• Novelty vs Preference
• Information Bubbles

Innovation

• HERec
• Information Cocoon Problem
• User-Adjustable Novelty Mechanism