Reward Guided Latent Consistency Distillation

arXiv:2403.11027v1 Announce Type: new
Abstract: Latent Consistency Distillation (LCD) has emerged as a promising paradigm for efficient text-to-image synthesis. By distilling a latent consistency model (LCM) from a pre-trained teacher latent diffusion model (LDM), LCD facilitates the generation of high-fidelity images within merely 2 to 4 inference steps. However, the LCM's efficient inference is obtained at the cost of the sample quality. In this paper, we propose compensating the quality loss by aligning LCM's output with human preference during training. …

abstract arxiv consistency model cs.ai cs.cv diffusion diffusion model distillation fidelity however image images inference ldm paradigm synthesis text text-to-image type