Skip to content
Research

MLIPS: Multi-Layer Indexer Pool Sharing.

arXiv:2610.27438v2 Announce Type: replace-cross Abstract: Sparse attention limits the keys used by each query, but finding those keys can still require a full-prefix scan at every layer. We introduce $MLIPS:$ $Multi$-$Layer$ $Indexer$ $Pool$ $Sharing$, a train

arXiv cs.AIResearch source·