Search papers, labs, and topics across Lattice.
Chinese Academy of Sciences, University of Chinese Academy of Sciences
3
0
5
Learned attention allocation patterns reveal that SWA is best positioned in lower layers, challenging conventional wisdom on attention distribution in LLMs.
Achieving up to 10x weight compression in LLMs without altering weights could revolutionize GPU memory usage and model deployment efficiency.
Graph-based nearest neighbor search, previously bottlenecked by memory bandwidth, achieves 20x speedup on billion-scale datasets using a novel processing-in-memory co-design.
