TailorMind: Towards Preference-Aligned Multimodal Content Generation

TailorMind is a novel framework for personalized multimodal content generation that integrates collaborative preference modeling with controllable generation techniques. It employs hypergraph collaborative filtering to enhance user profiles and utilizes retrieval-augmented style control to align outputs with user-generated content patterns, achieving improved coherence, novelty, and aesthetic quality compared to existing generation baselines. The accompanying TailorBench benchmark evaluates performance across five dimensions, with TailorMind demonstrating up to 29% gains in recall, making it a significant advancement for practitioners focused on generating user-tailored content without relying on existing datasets.