Learning User Preferences for Image Generation Model

2025-08-11

AI Summary

This work addresses the challenge of modeling fine-grained, dynamic, and heterogeneous user aesthetic preferences—spanning color, style, subject, composition, and other multi-level attributes—in personalized image generation, overcoming limitations of conventional approaches that rely on static user profiles or population-level average preferences. We propose a multimodal large language model–based framework for personalized preference learning. Our method introduces a contrastive preference loss and learnable user preference tokens to jointly model individual specificity and shared group patterns, enabling end-to-end learning of fine-grained preference representations from historical interaction data. Experiments demonstrate that our approach achieves significantly higher preference prediction accuracy than state-of-the-art baselines, enables high-fidelity clustering of aesthetically similar users, and substantially improves alignment between generated images and individual user preferences.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Humans and AI: Learning Human Values and Preferences
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

User preference prediction requires a comprehensive and accurate understanding of individual tastes. This includes both surface-level attributes, such as color and style, and deeper content-related aspects, such as themes and composition. However, existing methods typically rely on general human preferences or assume static user profiles, often neglecting individual variability and the dynamic, multifaceted nature of personal taste. To address these limitations, we propose an approach built upon Multimodal Large Language Models, introducing contrastive preference loss and preference tokens to learn personalized user preferences from historical interactions. The contrastive preference loss is designed to effectively distinguish between user ''likes'' and ''dislikes'', while the learnable preference tokens capture shared interest representations among existing users, enabling the model to activate group-specific preferences and enhance consistency across similar users. Extensive experiments demonstrate our model outperforms other methods in preference prediction accuracy, effectively identifying users with similar aesthetic inclinations and providing more precise guidance for generating images that align with individual tastes. The project page is exttt{https://learn-user-pref.github.io/}.

Problem

• Predicting dynamic user preferences for image generation
• Capturing multifaceted individual tastes accurately
• Enhancing consistency in personalized image generation

Innovation

• Multimodal Large Language Models for preference learning
• Contrastive preference loss distinguishes likes and dislikes
• Learnable preference tokens capture shared user interests