arXiv：cs.LG· Aksel Fristrup, Sumit Pandey, Ankit Kariryaa·· 4 小时前AI 评分39
ApexQuant：通过残差再各向同性化实现无数据弹性量化
ApexQuant: Data-Free Elastic Quantization by Residual Re-Isotropization
AI 导读
ApexQuant 是一种免校准的量化方法，通过递归重新量化残差误差，可作为现有量化器的精化层。该方法利用随机旋转使残差回归超球面均匀分布，从而在读取权重前确定目标误差所需的量化轮数，且每个前缀本身即为有效的低比特模型，一个产物可服务多种精度。研究在四个开源权重 LLM 及地球观测、医疗领域验证，4-bit 下接近全精度，并给出完全无数据设定下最优的 2-bit 方案。
正文
Abstract:We introduce ApexQuant, a calibration-free quantization method that recursively re-quantizes the residual error, serving as a refinement layer on top of existing quantizers. We establish that a fresh random rotation returns each residual to the uniform distribution on the hypersphere, which characterizes the rate of progressive error decay across successive passes. This result lets us determine, before any weight is read, how many passes a layer needs for a target weight-space error. Every prefix is itself a valid lower-rate model, so one artifact serves several precisions. We instantiate ApexQuant with three interchangeable stages, scalar, $E_8$ and trellis, and validate it on four open-weight LLMs and on Earth-observation and medical domains where in-distribution data is often unattainable as imagery arrives under restrictive licences or due to patient material under privacy constraints. Progressive re-isotropization comes within a few percent of full precision at four bits and gives the best two-bit arm we measure, in a completely data-free setting.
| Comments: | 25 pages, 6 figures |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.07904 [cs.LG] |
| (or arXiv:2610.07904v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07904 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Aksel Fristrup [view email]
[v1]
Tue, 6 Oct 2026 07:49:55 UTC (1,201 KB)
来源：arXiv：cs.LG · arxiv.org