Recommendations Beyond Catalogs: Diffusion Models for Personalized Generation

2025-02-05

arXiv.org

AI Summary

Existing diffusion-based image generation systems struggle to model implicit user feedback (e.g., ratings, clicks) and lack lightweight, scalable personalization capabilities. To address this, we propose the first diffusion framework for generative recommendation—departing from conventional retrieval-based paradigms that merely select from a fixed item catalog, our approach directly synthesizes novel preference-aligned images for individual users. We introduce an implicit diffusion prior mechanism grounded in users’ historical ratings, eliminating the need for explicit annotations. Furthermore, we design an embedding-space-efficient training strategy and a personalized decoding method. Critically, we establish the first evaluation metric suite specifically tailored for generative recommendation. Experiments on real-world data demonstrate substantial improvements in alignment between generated images and user preferences; under our new metrics, our method outperforms baselines by an average of 32.7%.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

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
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems

Abstract

Modern recommender systems follow the guiding principle of serving the right user, the right item at the right time. One of their main limitations is that they are typically limited to items already in the catalog. We propose REcommendations BEyond CAtalogs, REBECA, a new class of probabilistic diffusion-based recommender systems that synthesize new items tailored to individual tastes rather than retrieve items from the catalog. REBECA combines efficient training in embedding space with a novel diffusion prior that only requires users' past ratings of items. We evaluate REBECA on real-world data and propose novel personalization metrics for generative recommender systems. Extensive experiments demonstrate that REBECA produces high-quality, personalized recommendations, generating images that align with users' unique preferences.

Problem

• Addressing user diversity limitations in diffusion-based image generation systems
• Eliminating reliance on costly paired preference data for personalization
• Providing scalable personalized image generation without model fine-tuning

Innovation

• Lightweight framework using implicit feedback signals
• Two-stage process with conditional diffusion model
• Fine-tuning-free personalization for large user bases