Skip to content
Hot eventLive

Nucleus Speculative Decoding 论文发布

1 reports1 sources3 hr ago updated

Get the story

AI overview

2026年10月7日,研究者在 arXiv(Computation and Language)发布 Nucleus Speculative Decoding(NSD)论文。该方法在标准接受规则之外引入目标模型的 nucleus 判定,接受高可信度的草稿 token。论文给出单步误差与序列级保真界的理论分析,实验显示吞吐量最高达自回归解码的 5.16 倍、标准投机解码的 3.15 倍。代码已在 GitHub 开源。

Generated from reports · updated 3 hr ago

Timeline

Follow the coverage from different angles.

Oct 7, 2026
  1. arXiv · Computation and Language
    Nucleus Speculative Decoding:基于目标模型可信度的宽松验证方法

    研究者提出 Nucleus Speculative Decoding(NSD),在标准接受规则之外引入目标模型 nucleus 判定,接受高可信度的草稿 token。理论分析给出单步误差与序列级保真界,实验显示吞吐量最高达自回归解码的 5.16 倍、标准投机解码的 3.15 倍。代码已在 GitHub 开源。

Heat trend

Current heat 9·Comparable peak 10(Oct 7)·Comparable change over 24 hours –

02.557.510Oct7Oct7Oct7Oct7

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.