DIFFUSION-BASED COUNTERFACTUALS

Summary from the TLDR AI newsletter

This paper introduces two methods that use diffusion models to generate counterfactual explanations in image regression tasks, highlighting trade-offs in sparsity and quality between pixel and latent space approaches.