Search papers, labs, and topics across Lattice.
1
0
1
Video diffusion models hallucinate rich scene completions but inevitably drift; anchoring their appearance priors directly inside a 3D geometric latent space eliminates sequential inconsistency and slashes trajectory error by 32%.
