Unified Concept Editing in Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gandikota, Rohit, Orgad, Hadas, Belinkov, Yonatan, Materzyńska, Joanna, Bau, David
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916451141550080
author Gandikota, Rohit
Orgad, Hadas
Belinkov, Yonatan
Materzyńska, Joanna
Bau, David
author_facet Gandikota, Rohit
Orgad, Hadas
Belinkov, Yonatan
Materzyńska, Joanna
Bau, David
contents Text-to-image models suffer from various safety issues that may limit their suitability for deployment. Previous methods have separately addressed individual issues of bias, copyright, and offensive content in text-to-image models. However, in the real world, all of these issues appear simultaneously in the same model. We present a method that tackles all issues with a single approach. Our method, Unified Concept Editing (UCE), edits the model without training using a closed-form solution, and scales seamlessly to concurrent edits on text-conditional diffusion models. We demonstrate scalable simultaneous debiasing, style erasure, and content moderation by editing text-to-image projections, and we present extensive experiments demonstrating improved efficacy and scalability over prior work. Our code is available at https://unified.baulab.info
format Preprint
id arxiv_https___arxiv_org_abs_2308_14761
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Unified Concept Editing in Diffusion Models
Gandikota, Rohit
Orgad, Hadas
Belinkov, Yonatan
Materzyńska, Joanna
Bau, David
Computer Vision and Pattern Recognition
Machine Learning
Text-to-image models suffer from various safety issues that may limit their suitability for deployment. Previous methods have separately addressed individual issues of bias, copyright, and offensive content in text-to-image models. However, in the real world, all of these issues appear simultaneously in the same model. We present a method that tackles all issues with a single approach. Our method, Unified Concept Editing (UCE), edits the model without training using a closed-form solution, and scales seamlessly to concurrent edits on text-conditional diffusion models. We demonstrate scalable simultaneous debiasing, style erasure, and content moderation by editing text-to-image projections, and we present extensive experiments demonstrating improved efficacy and scalability over prior work. Our code is available at https://unified.baulab.info
title Unified Concept Editing in Diffusion Models
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2308.14761