Search papers, labs, and topics across Lattice.
1
0
3
Text-to-video models can now generate videos that actually respect spatial relationships, thanks to a new geometry-based reward function that beats VLM-based alternatives.
