Search papers, labs, and topics across Lattice.
1
0
3
6
Ditch the heuristic latent spaces: Geometric Autoencoders offer a principled way to inject VFM priors into diffusion models, yielding state-of-the-art image generation with better compression and semantic depth.
