提升NVFP4量化精度，通过联合优化缩放因子实现更准重建。
SOAR: Scale Optimization for Accurate Reconstruction in NVFP4 Quantization
- 联合优化全局与块级缩放因子，解析求解最小化重建误差。
- 解耦量化与反量化缩放，离散搜索减少精度损失。
- 无需额外硬件，适配大模型量化部署场景。
NVFP4是一种新兴的高效4比特微缩放格式，适用于大型语言模型（LLMs），具备优异的数值保真度并获得原生硬件支持。然而，现有方法因缩放因子选择僵化及量化与反量化缩放耦合处理，常导致性能不佳。为此，本文提出一种新的后训练量化框架SOAR，用于提升NVFP4量化精度。其核心是闭式联合缩放优化（CJSO），通过重建误差最小化推导出解析解，联合优化全局与块级缩放因子。此外，引入解耦缩放搜索（DSS），将高精度量化缩放与其受限的反量化缩放解耦，并进行离散搜索，以缓解缩放量化带来的精度损失。在多个LLM上的大量实验表明，本方法持续优于现有NVFP4量化基线，在相同内存开销下实现更高精度，且无额外硬件开销。代码与模型将开源于https://github.com/steven-bao1/SOAR。
原文摘要 · Abstract (English)
NVFP4 has recently emerged as an efficient 4-bit microscaling format for large language models (LLMs), offering superior numerical fidelity with native hardware support. However, existing methods often yield suboptimal performance due to inflexible scale selection and the coupled treatment of quantization and dequantization scales. To address these issues, we propose Scale Optimization for Accurate Reconstruction (SOAR), a novel post-training quantization framework that improves the accuracy of NVFP4 quantization. At its core, SOAR features Closed-form Joint Scale Optimization (CJSO), which jointly optimizes global and block-wise scales via analytical solutions derived from reconstruction error minimization. Furthermore, it incorporates Decoupled Scale Search (DSS). DSS decouples the high-precision quantization scale from its constrained dequantization counterpart, and performs discrete search to mitigate precision loss from scale quantization. Extensive experiments across multiple LLMs show that our method consistently outperforms existing NVFP4 quantization baselines, achieving superior accuracy under the same memory footprint with no additional hardware overhead. The code and models will be available at https://github.com/steven-bao1/SOAR.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。