GemiRec: Interest Quantization and Generation for Multi-Interest Recommendation

2025-10-16

AI Summary

Multi-interest recommendation faces two key challenges: interest collapse—where user representations become homogeneous—and insufficient modeling of interest evolution—hindering the capture of latent interests without explicit user interactions. To address these, we propose GemiRec, a novel framework comprising two core components: (1) a vector quantization mechanism that constructs a shared interest dictionary to structurally disentangle multiple interests and mitigate collapse; and (2) a generative interest evolution module that explicitly models the dynamic progression of latent interests, enhancing prediction of emerging interests. GemiRec integrates a dual-tower architecture, posterior distribution modeling, and multi-interest retrieval, while remaining compatible with industrial ranking systems. Theoretical analysis and extensive experiments demonstrate significant improvements in both recommendation diversity and accuracy. Deployed in production since March 2025, GemiRec has delivered stable and substantial performance gains in real-world scenarios.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Learning Preferences or Rankings
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Multi-interest recommendation has gained attention, especially in industrial retrieval stage. Unlike classical dual-tower methods, it generates multiple user representations instead of a single one to model comprehensive user interests. However, prior studies have identified two underlying limitations: the first is interest collapse, where multiple representations homogenize. The second is insufficient modeling of interest evolution, as they struggle to capture latent interests absent from a user's historical behavior. We begin with a thorough review of existing works in tackling these limitations. Then, we attempt to tackle these limitations from a new perspective. Specifically, we propose a framework-level refinement for multi-interest recommendation, named GemiRec. The proposed framework leverages interest quantization to enforce a structural interest separation and interest generation to learn the evolving dynamics of user interests explicitly. It comprises three modules: (a) Interest Dictionary Maintenance Module (IDMM) maintains a shared quantized interest dictionary. (b) Multi-Interest Posterior Distribution Module (MIPDM) employs a generative model to capture the distribution of user future interests. (c) Multi-Interest Retrieval Module (MIRM) retrieves items using multiple user-interest representations. Both theoretical and empirical analyses, as well as extensive experiments, demonstrate its advantages and effectiveness. Moreover, it has been deployed in production since March 2025, showing its practical value in industrial applications.

Problem

• Addresses interest collapse in multi-interest recommendation systems
• Models latent user interests absent from historical behavior
• Enhances interest separation and evolution through quantization and generation

Innovation

• Uses interest quantization for structural separation
• Employs generative model for interest distribution
• Retrieves items with multiple user representations