DivRL: Disentangled Self-Similarity Rewards for Diverse Subject-Driven Generation

arXiv:2606.23950v1 Announce Type: new Abstract: Subject-driven image generation faces an "Identity-Diversity Paradox", where strong identity preservation often leads to rigid and low-diversity outputs. We propose a post-training framework called DivRL that jointly optimizes identity consistency and structural diversity simultaneously by leveraging disentangled visual features from a robust similarity model. Specifically, we introduce a Negative Self-Similarity Measure (nSSM) to quantify structur...