Research · March 12, 2015
A generative process of noising and denoising was described.
It offered a probabilistic route to generating data through repeated denoising.
This early paper is an antecedent, not a claim about every later image generator.
arXiv · Published March 12, 2015
Both records concern generative image models; that paper is earlier work in the field, not the basis of this particular model.
Both describe image generation learned from data; the training mechanisms differ.
The same move: a generative model with an explicit probabilistic justification.
Auto-Encoding Variational BayesThe model belongs to the diffusion family.
High-Resolution Image Synthesis with Latent Diffusion ModelsVideo is generated by the same process of gradually removing noise.
Sora: Creating video from textThe structure is generated by a diffusion module in place of the earlier geometric block.
Accurate structure prediction of biomolecular interactions with AlphaFold 3Generating an image frame by frame, but each next one depends on the viewer's action.
Both outstanding papers work inside this family: one on order in diffusion language models, the other on the accuracy of sampling from diffusion models.
Announcing the ICML 2026 AwardsThe same forward and reverse diffusion described in 2015; this paper builds training on a weighted variational bound on top of it.
Denoising Diffusion Probabilistic Models