Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models

arXiv:2404.13706v1 Announce Type: cross
Abstract: Motivated by ethical and legal concerns, the scientific community is actively developing methods to limit the misuse of Text-to-Image diffusion models for reproducing copyrighted, violent, explicit, or personal information in the generated images. Simultaneously, researchers put these newly developed safety measures to the test by assuming the role of an adversary to find vulnerabilities and backdoors in them. We use compositional property of diffusion models, which allows to leverage multiple prompts in a single image …

abstract arxiv community concept concerns cs.ai cs.cv cs.lg diffusion diffusion models ethical generated image image diffusion images information legal misuse personal information researchers safety safety measures scientific test text text-to-image type