Search papers, labs, and topics across Lattice.
1
0
3
Cut sparse attention indexing costs by 75% without sacrificing quality by cleverly reusing top-k indices across layers.
