Search papers, labs, and topics across Lattice.
University of Hong Kong
1
0
3
Integrating visual frames with textual reasoning steps, VTI-CoT achieves state-of-the-art video reasoning performance while boosting training efficiency.
