Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
3
0
Distributed attention can achieve near-zero communication overhead for mask localization and compute-communication overlap, unlocking massive throughput gains for long-context LLMs.