Search papers, labs, and topics across Lattice.
Affiliation:
1
3
7
Current video generation models struggle with visual reasoning, with the best achieving only 51% accuracy on a new benchmark designed to probe their capabilities.
