Search papers, labs, and topics across Lattice.
1
0
3
Uniformly sampling frames in video LLMs is leaving crucial temporal information on the cutting room floor: GroundVTS selectively attends to the most informative segments, substantially boosting grounding performance.
