PulseAugur
中
实时 04:09:52
English(EN) RIFT: KV Cache Compression for Draft Models with Complementary Compressibility Training

RIFT 改进了 LLM 中投机解码的 KV 缓存压缩

作者 PulseAugur 编辑部 · [2 个来源] ·

研究人员推出了一种新颖的 KV 缓存压缩策略 RIFT，旨在提高大型语言模型中投机解码的效率。RIFT 解决了“解码过时”的问题，即压缩方法在长输出生成过程中会失去相关性。通过平衡预填充信息和最近生成的 token，并采用自适应刷新策略，RIFT 在管理解码延迟的同时提高了草稿模型的质量。这种方法结合了可压缩性驱动的训练，使得 Qwen3.5-4B 模型能够为 Qwen3.5-122B 进行草稿，同时在大量输入和输出上保留超过 98% 的全缓存接受长度。 AI

影响 提高了大型语言模型中长上下文生成的效率和成本效益。

排序理由 该集群描述了一篇详细介绍 LLM 中 KV 缓存压缩新方法的最新研究论文。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

RIFT 改进了 LLM 中投机解码的 KV 缓存压缩

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
Research
该集群描述了一篇详细介绍 LLM 中 KV 缓存压缩新方法的最新研究论文。
Source corroboration
2 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
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
4 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 [2]

  1. arXiv cs.CL TIER_1 English(EN) · Dor Tsur, Maor Ashkenazi, Maximilian Golub, Peter Hicks, Yaniv Galron, Talor Abramovich, Yoav Miron, Tiyasa Mitra, Bita Darvish Rouhani, Ran Zilberstein ·

    RIFT: 针对具有互补可压缩性训练的草稿模型的 KV 缓存压缩

    arXiv:2610.31098v1 Announce Type: new Abstract: Speculative decoding is a popular decode acceleration method. As such, it is particularly relevant to long-output generation, which requires an increasingly large KV cache. KV cache compression offers a way to constrain costs in lon…

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    RIFT：用于具有互补可压缩性训练的草稿模型的KV缓存压缩

    Speculative decoding is a popular decode acceleration method. As such, it is particularly relevant to long-output generation, which requires an increasingly large KV cache. KV cache compression offers a way to constrain costs in long-context drafting. However, compression methods…

相关实体

相关话题