FOCUS: FP4 Optimization via Coupled-Relaxation and Dual-Granularity Scaling

📄 arXiv: 2608.01847v1 📥 PDF

作者: Xianglong Yan, Hong Liu, Chengzhu Bao, Tianao Zhang, Guanghua Yu, Jianchen Zhu, Yulun Zhang

分类: cs.AI

发布日期: 2026-08-03

🔗 代码/项目: GITHUB

💡 一句话要点

提出FOCUS框架以优化FP4量化精度问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: FP4量化 耦合松弛 双粒度缩放 大型语言模型 后训练量化 硬件兼容性 优化策略

📋 核心要点

  1. 现有FP4量化方法将量化和反量化尺度紧密耦合，导致优化空间未被充分利用，难以保持高准确性。
  2. FOCUS框架通过耦合松弛缩放（CRS）和双粒度缩放（DGS）解耦量化尺度，允许更灵活的优化，适应局部权重分布。
  3. 实验结果显示，FOCUS在多个LLM家族和基准测试中实现了最先进的FP4准确性，且没有增加推理开销。

📝 摘要（中文）

大型语言模型（LLMs）在性能上表现出色，但由于其庞大的体积，部署成本高昂。FP4量化（如MXFP4和NVFP4格式）为此提供了有吸引力的解决方案，然而在FP4精度下保持准确性仍然困难。现有方法将量化和反量化尺度紧密耦合，限制了优化空间。本文提出FOCUS，一个后训练量化框架，通过耦合松弛和双粒度缩放实现FP4优化，显著提升了量化精度，同时不增加推理开销。

🔬 方法详解

问题定义：本文旨在解决FP4量化中量化和反量化尺度耦合导致的准确性下降问题。现有方法强制两者遵循硬件要求的离散低精度格式，限制了优化潜力。

核心思路：FOCUS框架通过耦合松弛缩放（CRS）和双粒度缩放（DGS）来解耦量化尺度，CRS引入可学习的全精度系数，DGS则在更细的子块粒度上优化量化尺度，从而实现更有效的优化。

技术框架：FOCUS的整体架构包括两个主要模块：耦合松弛缩放模块和双粒度缩放模块。耦合松弛模块负责放宽量化和反量化尺度的耦合关系，而双粒度模块则在子块级别进行更精细的量化尺度调整。

关键创新：FOCUS的主要创新在于通过耦合松弛缩放解耦量化和反量化尺度，允许更灵活的优化策略，与现有方法相比，显著提升了FP4精度。

关键设计：在设计中，FOCUS使用了可学习的全精度系数来调整量化尺度，并在双粒度缩放中引入了局部权重分布的适应性调整，确保了硬件兼容性和优化效果。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

FOCUS在多个大型语言模型和基准测试中实现了最先进的FP4准确性，具体表现为在MXFP4和NVFP4格式下的准确性提升，且未增加推理开销，展示了其在实际应用中的有效性和优势。

🎯 应用场景

FOCUS框架在大型语言模型的量化部署中具有广泛的应用潜力，尤其是在需要高效推理和低延迟的场景中，如智能助手、自动翻译和实时语音识别等。其优化策略可为未来的AI模型部署提供更高的准确性和更低的资源消耗。

📄 摘要（原文）

Large language models (LLMs) achieve remarkable performance but are expensive to deploy due to their enormous size. FP4 quantization, with formats such as MXFP4 and NVFP4, offers an appealing solution with native hardware support on modern accelerators. However, maintaining accuracy under FP4 precision remains difficult. A key bottleneck lies in scale optimization: existing methods tightly couple the quantization and dequantization scales, forcing both to conform to the discrete low-precision format required by hardware, such as E8M0 in MXFP4. Yet the quantization scale is never stored and need not obey this constraint, suggesting a significant untapped optimization space. In this work, we propose FOCUS, a post-training quantization framework with end-to-end scale learning for FP4 Optimization via Coupled-Relaxation and Dual-Granularity Scaling. Coupled-Relaxation Scaling (CRS) relaxes the tight coupling between quantization and dequantization scales with a learnable full-precision coefficient, enabling more effective optimization without breaking hardware compliance. Dual-Granularity Scaling (DGS) further refines the quantization scale at a finer sub-block granularity, allowing more precise adaptation to local weight distributions. Experiments across multiple LLM families and benchmarks show that FOCUS achieves state-of-the-art FP4 accuracy under both MXFP4 and NVFP4 formats, while introducing no additional inference overhead. Code and quantized models will be released at https://github.com/tencent/AngelSlim.