Circumventing Concept Erasure Methods For Text-to-Image Generative Models
Text-to-image generative models can produce photo-realistic images for an extremely broad range of concepts, and their usage has proliferated widely among the general public. On the flip side, these models have numerous drawbacks, including their potential to generate images featuring sexually explicit content, mirror artistic styles without permission, or even hallucinate (or deepfake) the likenesses of celebrities. Consequently, various methods have been proposed in order to "erase" sensitive concepts from text-to-image models. In this work, we examine five recently proposed concept erasure methods, and show that targeted concepts are not fully excised from any of these methods. Specifically, we leverage the existence of special learned word embeddings that can retrieve "erased" concepts from the sanitized models with no alterations to their weights. Our results highlight the brittleness of post hoc concept erasure methods, and call into question their use in the algorithmic toolkit for AI safety.
Code (1)
Tasks
Face SwappingWord EmbeddingsMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Dark Miner: Defend against undesired generation for text-to-image diffusion models
Text-to-image diffusion models have been demonstrated with undesired generation due to unfiltered large-scale training data, such as sexual images and copyrights, necessitating the erasure of undesired concepts. Most exi…
CRCE: Coreference-Retention Concept Erasure in Text-to-Image Diffusion Models
Text-to-Image diffusion models can produce undesirable content that necessitates concept erasure. However, existing methods struggle with under-erasure, leaving residual traces of targeted concepts, or over-erasure, mist…
Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration
Text-to-image generative models have achieved impressive fidelity and diversity, but can inadvertently produce unsafe or undesirable content due to implicit biases embedded in large-scale training datasets. Existing conc…
Text-to-Image GenerationRepresentation LearningM-ErasureBench: A Comprehensive Multimodal Evaluation Benchmark for Concept Erasure in Diffusion Models
Text-to-image diffusion models may generate harmful or copyrighted content, motivating research on concept erasure. However, existing approaches primarily focus on erasing concepts from text prompts, overlooking other in…
Image EditingACE: Anti-Editing Concept Erasure in Text-to-Image Models
Recent advance in text-to-image diffusion models have significantly facilitated the generation of high-quality images, but also raising concerns about the illegal creation of harmful content, such as copyrighted images. …