EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders
Paper • 2609.03629 • Published
EraseSAE Sparse Autoencoder Checkpoints
This repository contains the sparse autoencoder checkpoints and configuration files from the paper EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders.
These checkpoints are intended to be used with the EraseSAE framework for concept erasure in text-to-video diffusion models, supporting tasks such as nudity and celebrity identity erasure.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support