MACE: Mass Concept Erasure in Diffusion Models

TL;DR

MACE is a finetuning framework that effectively erases up to 100 concepts in diffusion models, balancing generality and specificity to prevent unwanted content generation.

Contribution

MACE introduces a scalable concept erasure method using cross-attention refinement and LoRA finetuning, surpassing existing methods in erasing multiple concepts simultaneously.

Findings

Outperforms prior methods in object, celebrity, explicit content, and artistic style erasure.

Successfully erases up to 100 concepts while maintaining unrelated content.

Achieves effective balance between erasing synonyms and preserving unrelated concepts.

Abstract

The rapid expansion of large-scale text-to-image diffusion models has raised growing concerns regarding their potential misuse in creating harmful or misleading content. In this paper, we introduce MACE, a finetuning framework for the task of mass concept erasure. This task aims to prevent models from generating images that embody unwanted concepts when prompted. Existing concept erasure methods are typically restricted to handling fewer than five concepts simultaneously and struggle to find a balance between erasing concept synonyms (generality) and maintaining unrelated concepts (specificity). In contrast, MACE differs by successfully scaling the erasure scope up to 100 concepts and by achieving an effective balance between generality and specificity. This is achieved by leveraging closed-form cross-attention refinement along with LoRA finetuning, collectively eliminating the information of undesirable concepts. Furthermore, MACE integrates multiple LoRAs without mutual interference. We conduct extensive evaluations of MACE against prior methods across four different tasks: object erasure, celebrity erasure, explicit content erasure, and artistic style erasure. Our results reveal that MACE surpasses prior methods in all evaluated tasks. Code is available at https://github.com/Shilin-LU/MACE.