Search papers, labs, and topics across Lattice.
1
0
3
Video-LLMs can perfectly label every object in a scene yet collapse to near-random chance when simply asked which of two identical actions moved faster.
