Skip to content
Hot storiesWatching

研究团队提出MXCode超低比特量化框架

1 reports1 reporting sourcesUpdated 1 day ago

Start here

AI overview

2026年11月3日,arXiv机器学习领域发布了一项关于MXCode矢量量化框架的研究。该框架面向大规模MX原生推理,将超低比特矢量量化(VQ)与MX原生FP4计算相结合,无需物化中间高精度权重即可实现单权重约2-bit存储。MXCode框架涵盖PostScale、PreScale与PreScale+三种策略,并已在Qwen3(235B)、DeepSeek-R1(671B)和GLM-5.2(753B)等超大规模模型上完成了验证。

AI-generated from reports · Updated 23 hours ago

Report timeline

Follow the reports to see different sides of the story.

Nov 3
  1. arXiv 机器学习
    MXCode:面向大规模 MX 原生推理的矢量量化框架

    MXCode 框架将超低比特矢量量化(VQ)与 MX 原生 FP4 计算结合,无需物化中间高精度权重即可实现单权重约 2-bit 存储。该框架包含 PostScale、PreScale 与 PreScale+ 三种策略,并在 Qwen3 (235B)、DeepSeek-R1 (671B) 和 GLM-5.2 (753B) 上完成验证。

Story heat trend

Current heat 5·Comparable peak 9(Nov 3 18:00)·Comparable change, last 24 hours–

02.557.510Nov3Nov4Nov4Nov4

Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.