The preprint notes that visual generators can produce photorealistic frames without preserving a consistent 3D scene. It then presents a compact geometry-native latent space as a shared foundation for perception and generation.
The paper is titled GAE, and the name 'geometry-native' suggests the latent space is designed around geometric structure. Because the abstract is truncated, the full argument and any experimental validation are not available in the source text.
This digest is based on a single source, so there are no conflicting findings to note. The abstract does not include experimental results or comparisons with existing methods.