FairGen: Enhancing Fairness in Text-to-Image Diffusion Models via Self-Discovering Latent Directions

TL;DR

FairGen is a novel, lightweight, plug-and-play method that learns attribute directions in latent space without reference datasets, effectively reducing biases in text-to-image diffusion models across multiple attributes.

Contribution

It introduces a self-discovering approach to learn attribute latent directions, enabling bias mitigation without retraining or reference datasets.

Findings

Outperforms previous state-of-the-art methods in debiasing.

Effectively reduces gender and racial biases in generated images.

Simultaneously debiases multiple attributes with high efficiency.

Abstract

While Diffusion Models (DM) exhibit remarkable performance across various image generative tasks, they nonetheless reflect the inherent bias presented in the training set. As DMs are now widely used in real-world applications, these biases could perpetuate a distorted worldview and hinder opportunities for minority groups. Existing methods on debiasing DMs usually requires model retraining with a human-crafted reference dataset or additional classifiers, which suffer from two major limitations: (1) collecting reference datasets causes expensive annotation cost; (2) the debiasing performance is heavily constrained by the quality of the reference dataset or the additional classifier. To address the above limitations, we propose FairGen, a plug-and-play method that learns attribute latent directions in a self-discovering manner, thus eliminating the reliance on such reference dataset. Specifically, FairGen consists of two parts: a set of attribute adapters and a distribution indicator. Each adapter in the set aims to learn an attribute latent direction, and is optimized via noise composition through a self-discovering process. Then, the distribution indicator is multiplied by the set of adapters to guide the generation process towards the prescribed distribution. Our method enables debiasing multiple attributes in DMs simultaneously, while remaining lightweight and easily integrable with other DMs, eliminating the need for retraining. Extensive experiments on debiasing gender, racial, and their intersectional biases show that our method outperforms previous SOTA by a large margin.