This AI Paper Proposes A New Method For Fine-Tuning Model Weights To Erase Concepts From Diffusion Models Using Their Own Knowledge

The Stable Diffusion text-to-image diffusion model has been released as open-source. They provide a method for selecting and eliminating a single idea from the weights of a pretrained text-conditional model. They use just undesirable concept descriptions and no further training data to fine-tune the model’s parameters.