Search papers, labs, and topics across Lattice.
1
0
3
Sub-linear attention is now possible without sacrificing complete long-range dependency retention, thanks to learnable summary tokens that compress context.
