LATO.2 introduces a factorized flow-matching framework for topology-aware 3D mesh generation. It separates vertex generation from connectivity generation, conditioning the latter on realized vertices, while both stages share a coarse voxel scaffold. Dedicated VAEs recover vertices at sub-voxel precision and encode discrete connectivity into continuous latents. The authors describe two capabilities: part-wise synthesis at higher resolution than a monolithic latent allows, and topology-adaptive editing where vertex manipulation induces corresponding connectivity without re-optimization. The abstract reports improvements over state-of-the-art topology-aware mesh generators in geometric fidelity and connectivity quality, but provides no quantitative results or benchmark details.
No heat snapshots are available in the last 24 hours.