PapersBuilders13:30Chinese contentLatent Sparse Attention:让超长上下文注意力计算提速 72 倍超长上下文跑不动?这篇论文用分层聚类做稀疏注意力,10M token 下比全注意力快 72 倍还不掉点,做长上下文的值得看看。#Latent Sparse Attention#GLM-5.3-FLASH#稀疏注意力#长上下文aarXiv cs.AI@Ales Kubicek et al. (10)SourceRead laterReadWorth followingUsefulFollow Latent Sparse Attention