Self-Discovering Interpretable Diffusion Latent Directions for Responsible Text-to-Image Generation

TL;DR

This paper introduces a self-supervised method to discover interpretable latent directions in diffusion models, enabling better understanding and mitigation of biased or harmful content in text-to-image generation.

Contribution

It proposes a novel self-supervised approach to identify latent directions for arbitrary concepts, including inappropriate ones, and uses this for responsible content mitigation.

Findings

Effective in fair and safe image generation

Able to discover latent directions for arbitrary concepts

Improves responsible text-to-image synthesis

Abstract

Diffusion-based models have gained significant popularity for text-to-image generation due to their exceptional image-generation capabilities. A risk with these models is the potential generation of inappropriate content, such as biased or harmful images. However, the underlying reasons for generating such undesired content from the perspective of the diffusion model's internal representation remain unclear. Previous work interprets vectors in an interpretable latent space of diffusion models as semantic concepts. However, existing approaches cannot discover directions for arbitrary concepts, such as those related to inappropriate concepts. In this work, we propose a novel self-supervised approach to find interpretable latent directions for a given concept. With the discovered vectors, we further propose a simple approach to mitigate inappropriate generation. Extensive experiments have been conducted to verify the effectiveness of our mitigation approach, namely, for fair generation, safe generation, and responsible text-enhancing generation. Project page: \url{https://interpretdiffusion.github.io}.