Search papers, labs, and topics across Lattice.
1
0
3
5
Current video generation models struggle with visual reasoning, achieving only 51% accuracy on a new benchmark designed to probe their capabilities.
