Diffusion-RPO: Aligning Diffusion Models through Relative Preference Optimization

TL;DR

This paper introduces Diffusion-RPO, a novel method for aligning diffusion-based text-to-image models with human preferences using relative preference optimization and a new style alignment metric, outperforming previous techniques.

Contribution

The paper presents Diffusion-RPO, a new approach that improves preference alignment in diffusion models by utilizing cross-modal pairs and a novel evaluation metric.

Findings

Diffusion-RPO outperforms Supervised Fine-Tuning and Diffusion-DPO.

It achieves better human preference alignment in automated evaluations.

The style alignment metric provides more interpretable assessment of model preferences.

Abstract

Aligning large language models with human preferences has emerged as a critical focus in language modeling research. Yet, integrating preference learning into Text-to-Image (T2I) generative models is still relatively uncharted territory. The Diffusion-DPO technique made initial strides by employing pairwise preference learning in diffusion models tailored for specific text prompts. We introduce Diffusion-RPO, a new method designed to align diffusion-based T2I models with human preferences more effectively. This approach leverages both prompt-image pairs with identical prompts and those with semantically related content across various modalities. Furthermore, we have developed a new evaluation metric, style alignment, aimed at overcoming the challenges of high costs, low reproducibility, and limited interpretability prevalent in current evaluations of human preference alignment. Our findings demonstrate that Diffusion-RPO outperforms established methods such as Supervised Fine-Tuning and Diffusion-DPO in tuning Stable Diffusion versions 1.5 and XL-1.0, achieving superior results in both automated evaluations of human preferences and style alignment. Our code is available at https://github.com/yigu1008/Diffusion-RPO