Search papers, labs, and topics across Lattice.
1
0
3
Misalignment between visual evidence and predicted timestamps in video grounding can lead to substantial performance drops, but CAVE effectively bridges this gap with boundary-specific rewards.