Diffusion Model for Interest Refinement in Multi-Interest Recommendation

2025-02-08

AI Summary

Existing attention-based multi-interest recommendation methods model only item-level relevance, yielding coarse, noisy, and poorly discriminative interest representations. To address this, we propose the Dimension-level Interest Refinement Framework (DIRF), the first to integrate diffusion models into multi-interest recommendation. DIRF injects controllable noise into the embedding dimension space and employs cross-attention mechanisms alongside dynamic item pruning to enable collaboratively guided iterative reconstruction—thereby achieving fine-grained interest purification. This refines interest representations to enhance both discriminability and purity. Offline experiments demonstrate consistent superiority over state-of-the-art methods across multiple benchmarks. Online A/B tests on a large-scale production system—serving hundreds of millions of daily active users—show statistically significant improvements in click-through rate and user satisfaction. DIRF has been successfully deployed in this real-world recommender system.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Learning Preferences or Rankings
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Multi-interest candidate matching plays a pivotal role in personalized recommender systems, as it captures diverse user interests from their historical behaviors. Most existing methods utilize attention mechanisms to generate interest representations by aggregating historical item embeddings. However, these methods only capture overall item-level relevance, leading to coarse-grained interest representations that include irrelevant information. To address this issue, we propose the Diffusion Multi-Interest model (DMI), a novel framework for refining user interest representations at the dimension level. Specifically, DMI first introduces controllable noise into coarse-grained interest representations at the dimensional level. Then, in the iterative reconstruction process, DMI combines a cross-attention mechanism and an item pruning strategy to reconstruct the personalized interest vectors with the guidance of tailored collaborative information. Extensive experiments demonstrate the effectiveness of DMI, surpassing state-of-the-art methods on offline evaluations and an online A/B test. Successfully deployed in the real-world recommender system, DMI effectively enhances user satisfaction and system performance at scale, serving the major traffic of hundreds of millions of daily active users. footnote{The code will be released for reproducibility once the paper is accepted.}

Problem

• Refines user interest representations in recommendations
• Addresses coarse-grained interest representations issue
• Enhances system performance and user satisfaction

Innovation

• Diffusion model refines user interests
• Cross-attention and pruning reconstruct vectors
• Deployed in large-scale recommender systems