English(EN) Long-Context Drafting Does Not Require Long-Context Drafters

研究：短上下文起草器无需再训练即可处理200万个标记

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-16 04:00

一篇新的研究论文提出，语言模型中的长上下文起草不需要专门的长上下文起草器。作者认为，通过冻结现有的短上下文起草器，并在推理时使用机制来管理其注意力缓存和重新定位索引，它们可以有效地将其上下文窗口扩展到两百万个标记。据报道，这种方法在各种目标起草器对和语料库中都保持了投机解码的速度提升优势，而无需对起草器模型进行任何权重修改。 AI

影响这项研究可以显著降低处理LLM中极长上下文的计算成本和复杂性。

排序理由该集群包含一篇详细介绍语言模型推理新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 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: 16 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍语言模型推理新方法的论文。[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: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Aref Jafari, Vansh Bhatia, Matt Qin, Subhajit Dutta Chowdhury, Ben Lee, Vikram Appia, Ali Ghodsi, Mehdi Rezagholizadeh · 2026-10-16 04:00

长上下文起草无需长上下文起草者

arXiv:2610.20865v1 Announce Type: new Abstract: Speculative decoding pairs a cheap drafter with parallel verification against the target. At ultra-long context it stops paying, because a drafter that attends over the full prefix inherits the target's context-dependent cost, so th…

本文如何被排名

报道来源 [1]

长上下文起草无需长上下文起草者

相关实体

相关话题