Unsupervised learning of disentangled representations via diffusion variational autoencoders

ICLR 2026 Conference Submission13779 Authors

Keywords: unsupervised learning, diffusion models, variational autoencoders, disentangling, ELBO

Abstract: We present the diffusion variational autoencoder (DiVA), a method for unsupervised representation learning that combines the theoretical frameworks of diffusion models and VAEs. By unifying their respective evidence lower bounds, DiVA formulates a principled objective that learns representations through score-based guidance of the underlying diffusion process. The resulting representations automatically capture meaningful structure in the data: it recovers ground truth generative factors in synthetic datasets, learns factorized, semantic latent dimensions from complex natural images, and encodes video sequences into latent trajectories that are straighter than those of alternative encoders, despite training exclusively on static images. Furthermore, DiVA can extract useful representations from pre-trained diffusion models with minimal additional training. Finally, the explicitly probabilistic formulation provides new ways to identify semantically meaningful axes in the absence of supervised labels. Overall, these results indicate that implicit structural information in diffusion models can be made explicit and interpretable through synergistic combination with a variational autoencoder.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 13779