Large Language Models Enhanced Hyperbolic Space Recommender Systems

2025-04-08

AI Summary

Existing recommender systems operate in Euclidean space, limiting their ability to capture hierarchical structures inherent in textual and semantic data, thereby constraining the fidelity of user preference modeling. To address this, we propose HyperLLM—a model-agnostic framework that pioneers the deep integration of large language models (LLMs) with hyperbolic geometry. Its core innovations are: (1) a structure-semantic dual-path hierarchical modeling mechanism—leveraging LLMs to generate multi-level categorical labels for structural hierarchy alignment, and introducing a meta-optimization strategy to bridge LLM-derived semantic embeddings with hyperbolic collaborative representations; and (2) support for contrastive learning and multi-level label-driven hierarchical alignment. Extensive experiments on multiple benchmark datasets demonstrate that HyperLLM outperforms state-of-the-art hyperbolic and LLM-based recommenders by over 40%, while significantly improving training stability and generalization capability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Recommender Systems
• Natural Language Processing: (Large) Language Models

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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) have attracted significant attention in recommender systems for their excellent world knowledge capabilities. However, existing methods that rely on Euclidean space struggle to capture the rich hierarchical information inherent in textual and semantic data, which is essential for capturing user preferences. The geometric properties of hyperbolic space offer a promising solution to address this issue. Nevertheless, integrating LLMs-based methods with hyperbolic space to effectively extract and incorporate diverse hierarchical information is non-trivial. To this end, we propose a model-agnostic framework, named HyperLLM, which extracts and integrates hierarchical information from both structural and semantic perspectives. Structurally, HyperLLM uses LLMs to generate multi-level classification tags with hierarchical parent-child relationships for each item. Then, tag-item and user-item interactions are jointly learned and aligned through contrastive learning, thereby providing the model with clear hierarchical information. Semantically, HyperLLM introduces a novel meta-optimized strategy to extract hierarchical information from semantic embeddings and bridge the gap between the semantic and collaborative spaces for seamless integration. Extensive experiments show that HyperLLM significantly outperforms recommender systems based on hyperbolic space and LLMs, achieving performance improvements of over 40%. Furthermore, HyperLLM not only improves recommender performance but also enhances training stability, highlighting the critical role of hierarchical information in recommender systems.

Problem

• Enhancing recommender systems with hierarchical information extraction
• Integrating LLMs and hyperbolic space for better user preference modeling
• Bridging semantic and collaborative spaces for improved recommendation accuracy

Innovation

• LLMs generate hierarchical tags for items
• Contrastive learning aligns tag-item interactions
• Meta-optimized strategy bridges semantic spaces