Search papers, labs, and topics across Lattice.
1
3
Over half of current video understanding benchmarks can be solved without any visual context, exposing a critical flaw in model evaluation.
