SubQ 1.1 Card: Linear-scaling sparse attention with 98% retrieval at 12M tokens [pdf]
mitchwainer
·
2026-06-16
·
via HN's home page
SubQ 1.1 Card: Linear-scaling sparse attention with 98% retrieval at 12M tokens [pdf]
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。