全为1比特：迈向大语言模型真正的1比特训练后量化
All for 1-Bit: Towards Genuine 1-Bit Post-Training Quantization for LLMs
浏览论文内容
中文总结 AI 辅助
提出AF1框架，通过零空间感知二元分解和层次Shapley分配，在严格1.0-BPW预算下实现大语言模型真正的1比特训练后量化，显著降低内存并加速推理。
中文摘要 AI 辅助
大语言模型（LLMs）取得了显著进展，但其巨大的存储和内存带宽需求仍然阻碍着高效部署。权重二值化是一种有前景的解决方案，但现有的基于二值化的训练后量化（PTQ）方法通常因隐藏开销而远超名义上的1比特存储目标。为解决这一差距，我们提出了All for 1-Bit（AF1），一个真正的大语言模型1比特PTQ框架。AF1包含两个互补组件：（1）零空间感知二元分解（NABF），通过Hessian感知的替代重参数化、零空间感知的二元分解和仅缩放的全局限定重建来改进二元重建；（2）层次Shapley分配（HiSA），利用层次Shapley敏感性分配结构容量。两者共同在PTQ设置中严格的1.0-BPW预算下保持模型精度。在LLaMA、Qwen和Gemma系列上的实验表明，AF1在困惑度和零样本准确率上始终优于现有的基于二值化的PTQ方法。与BF16相比，AF1在评估模型上平均实现了2.5倍的推理加速和超过90%的内存减少，为可部署的真正1比特大语言模型压缩提供了实用路径。用于复现的代码可在该https URL获取。
英文摘要
Large language models (LLMs) have achieved remarkable progress, yet their massive storage and memory-bandwidth demands still hinder efficient deployment. Weight binarization is a promising solution, but existing binarization-based post-training quantization (PTQ) methods usually far exceed the nominal 1-bit storage target due to hidden overhead. To address this gap, we propose All for 1-Bit (AF1), a genuine 1-bit PTQ framework for LLMs. AF1 comprises two complementary components: (1) Null-space-Aware Binary Factorization (NABF) for improving binary reconstruction through Hessian-aware surrogate reparameterization, null-space-aware binary factorization, and scale-only global reconstruction; and (2) Hierarchical Shapley Allocation (HiSA) for assigning structural capacity using hierarchical Shapley sensitivity. Together, they preserve model accuracy under a strict 1.0-BPW budget in the PTQ setting. Experiments on LLaMA, Qwen, and Gemma families show that AF1 consistently outperforms existing binarization-based PTQ methods in perplexity and zero-shot accuracy. Compared with BF16, AF1 achieves an average 2.5 times inference speedup and over 90% memory reduction across evaluated models, providing a practical path toward deployable genuine 1-bit compression for LLMs. The code for reproducibility is available at https://github.com/Kishon-zzx/AF1.
发表机构
- Houmo AI（后摩智能）
- School of Integrated Circuits, Peking University（北京大学集成电路学院）
机构由 AI 辅助整理，请以论文原文为准。
