Causal Diffusion Autoencoders: A Framework for Counterfactual Image Generation using Diffusion Probabilistic Models
CausalDiffAE, a diffusion-based framework for learning disentangled causal representations and enabling controllable counterfactual generation.