ConceptPrune: Concept Editing in Diffusion Models via Skilled Neuron Pruning

TL;DR

ConceptPrune is a training-free method that efficiently unlearns undesirable concepts from diffusion models by pruning a tiny fraction of weights, enhancing safety and bias mitigation.

Contribution

It introduces a novel, simple weight pruning technique for concept unlearning in diffusion models without additional training or fine-tuning.

Findings

Erases target concepts by pruning about 0.12% of weights

Effective across diverse concepts like styles, nudity, and bias

Robust against adversarial attacks

Abstract

While large-scale text-to-image diffusion models have demonstrated impressive image-generation capabilities, there are significant concerns about their potential misuse for generating unsafe content, violating copyright, and perpetuating societal biases. Recently, the text-to-image generation community has begun addressing these concerns by editing or unlearning undesired concepts from pre-trained models. However, these methods often involve data-intensive and inefficient fine-tuning or utilize various forms of token remapping, rendering them susceptible to adversarial jailbreaks. In this paper, we present a simple and effective training-free approach, ConceptPrune, wherein we first identify critical regions within pre-trained models responsible for generating undesirable concepts, thereby facilitating straightforward concept unlearning via weight pruning. Experiments across a range of concepts including artistic styles, nudity, object erasure, and gender debiasing demonstrate that target concepts can be efficiently erased by pruning a tiny fraction, approximately 0.12% of total weights, enabling multi-concept erasure and robustness against various white-box and black-box adversarial attacks.