PSR: Subject-Consistency Rewards for Multi-Subject Personalized Generation

Shulei Wang, Longhui Wei, XIN HE, Jianbo Ouyang, Hui Lu, Zhou Zhao, Qi Tian

Keywords: Personalized Generation

Abstract: Personalized generation models for a single subject have demonstrated remarkable effectiveness, highlighting their significant potential. However, when extended to multiple subjects, existing models often exhibit degraded performance, particularly in maintaining subject consistency and adhering to textual prompt. We attribute these limitations to the absence of high-quality multi-subject datasets and the lack of refined post-training strategies. To address these challenges, we construct a scalable multi-subject data generation pipeline, which leverages strong single-subject models to synthesize multi-subject training data. Using this dataset, we first enable single-subject personalization models to acquire knowledge of multi-image and multi-subject scenarios. Furthermore, to enhance both subject consistency and text controllability, we design a set of pairwise subject-consistency rewards and general-purpose rewards, which are incorporated into a refined reinforcement learning stage. To comprehensively evaluate multi-subject personalization, we introduce a new benchmark that assesses model performance using seven subsets across three dimensions. Extensive experiments demonstrate the effectiveness of our approach in advancing multi-subject personalized image generation.

Primary Area: generative models

Submission Number: 8765