Search papers, labs, and topics across Lattice.
University of Michigan
1
0
2
DLA adapts memory management in linear attention, preserving crucial information while reducing error accumulation over long sequences.
