Comprehensive Evaluation and Analysis for NSFW Concept Erasure in Text-to-Image Diffusion Models

TL;DR

This paper systematically evaluates NSFW concept erasure methods in text-to-image diffusion models, providing insights and practical guidance to improve content safety and prevent harmful outputs.

Contribution

It introduces a comprehensive toolkit and conducts the first systematic study of NSFW concept erasure methods in diffusion models.

Findings

Effective evaluation framework for NSFW concept erasure

Insights into mechanisms affecting erasure effectiveness

Guidelines for applying erasure methods in real-world scenarios

Abstract

Text-to-image diffusion models have gained widespread application across various domains, demonstrating remarkable creative potential. However, the strong generalization capabilities of diffusion models can inadvertently lead to the generation of not-safe-for-work (NSFW) content, posing significant risks to their safe deployment. While several concept erasure methods have been proposed to mitigate the issue associated with NSFW content, a comprehensive evaluation of their effectiveness across various scenarios remains absent. To bridge this gap, we introduce a full-pipeline toolkit specifically designed for concept erasure and conduct the first systematic study of NSFW concept erasure methods. By examining the interplay between the underlying mechanisms and empirical observations, we provide in-depth insights and practical guidance for the effective application of concept erasure methods in various real-world scenarios, with the aim of advancing the understanding of content safety in diffusion models and establishing a solid foundation for future research and development in this critical area.