新的LLA技术大幅缩减Transformer缓存大小，提升容量

作者 PulseAugur 编辑部 · [1 个来源] · 2026-07-20 04:00

研究人员开发了循环潜在注意力（LLA）技术，这是一种用于循环、权重绑定的Transformer的新型缓存压缩技术。LLA利用了跨递归步骤的键/值（K/V）缓存的结构化、低秩性质，存储紧凑的潜在向量而不是完整的K/V向量。该方法显著提高了批处理容量并减少了内存使用，使Ouro-1.4B等模型能够在单个H200 GPU上以4k上下文处理768个序列，压缩比达到21.3倍。LLA在与其他缓存压缩方法相比时表现出优越的性能，并在Huginn-3.5B等模型上显示出有希望的结果。 AI

影响减少大型语言模型的内存占用，可能支持更广泛的部署和更长的上下文窗口。

排序理由详细介绍Transformer架构新技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍Transformer架构新技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 58 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · James O' Neill, Fergal Reid · 2026-07-20 04:00

Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers

arXiv:2607.15456v1 Announce Type: cross Abstract: Looped, weight-tied Transformers reduce parameters by reusing a block, but decoding still stores a separate K/V cache for every recurrence step. We show that this loop-indexed cache is highly structured. For a fixed token, layer a…

本文如何被排名

报道来源 [1]

Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers

相关实体

相关话题