Diff-Instruct++: Training One-step Text-to-image Generator Model to Align with Human Preferences

TL;DR

Diff-Instruct++ is a novel, fast-converging method for aligning one-step text-to-image generators with human preferences, leveraging reinforcement learning principles and theoretical insights to outperform existing models.

Contribution

The paper introduces Diff-Instruct++, the first image data-free, human preference alignment method for one-step text-to-image models, with new theoretical understanding of CFG and RLHF connections.

Findings

Achieves high aesthetic and human preference scores on COCO dataset

Outperforms existing open-source models in human preference evaluation

Provides theoretical insights linking CFG with RLHF in diffusion models

Abstract

One-step text-to-image generator models offer advantages such as swift inference efficiency, flexible architectures, and state-of-the-art generation performance. In this paper, we study the problem of aligning one-step generator models with human preferences for the first time. Inspired by the success of reinforcement learning using human feedback (RLHF), we formulate the alignment problem as maximizing expected human reward functions while adding an Integral Kullback-Leibler divergence term to prevent the generator from diverging. By overcoming technical challenges, we introduce Diff-Instruct++ (DI++), the first, fast-converging and image data-free human preference alignment method for one-step text-to-image generators. We also introduce novel theoretical insights, showing that using CFG for diffusion distillation is secretly doing RLHF with DI++. Such an interesting finding brings understanding and potential contributions to future research involving CFG. In the experiment sections, we align both UNet-based and DiT-based one-step generators using DI++, which use the Stable Diffusion 1.5 and the PixelArt-$\alpha$ as the reference diffusion processes. The resulting DiT-based one-step text-to-image model achieves a strong Aesthetic Score of 6.19 and an Image Reward of 1.24 on the COCO validation prompt dataset. It also achieves a leading Human preference Score (HPSv2.0) of 28.48, outperforming other open-sourced models such as Stable Diffusion XL, DMD2, SD-Turbo, as well as PixelArt-$\alpha$. Both theoretical contributions and empirical evidence indicate that DI++ is a strong human-preference alignment approach for one-step text-to-image models. The homepage of the paper is https://github.com/pkulwj1994/diff_instruct_pp.