Search papers, labs, and topics across Lattice.
National University of Defense Technology
1
0
2
2
Open-weight Omni models struggle with binding accuracy, achieving only 41.55% on a new counterfactual benchmark, highlighting a critical gap in long-video comprehension.
