The Dual Power of Interpretable Token Embeddings: Jailbreaking Attacks and Defenses for Diffusion Model Unlearning
Paper • 2504.21307 • Published
SalUn Stable Diffusion v1.4 checkpoints
This repository mirrors converted SalUn UNet checkpoints from
siyich/ijd-unlearned-checkpoints.
The checkpoints use Stable Diffusion v1.4 as the base model and were produced by
the SalUn method from
"SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation".
They are stored as Diffusers UNet state dictionaries; use the remaining pipeline
components from CompVis/stable-diffusion-v1-4.
Included targets:
- Nudity
- Church
- Garbage Truck
- Parachute
- Tench
These files were converted and collected by the authors of "The Dual Power of Interpretable Token Embeddings"; they are not byte-identical to the original full Stable Diffusion checkpoints.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support