超越均匀比特分配：面向Matryoshka嵌入的可变量化方案

AI 摘要

该研究提出了一种可变比特分配框架，通过将嵌入向量划分为连续桶并采用贪心策略非均匀分配存储，以在固定内存预算下提升量化质量。实验表明，在具有Matryoshka属性的嵌入上，非均匀分配在相同压缩率下比均匀分配在PQ和SQ上分别提升最多8%和18%的召回率。该工作为大规模检索系统的结构感知压缩与索引提供了新方向。

正文节选

Quantization Beyond Uniform Bit Allocation Abstract. Quantization is a fundamental technique to handle the growing sizes of embeddings generated by modern models. Existing quantization schemes are largely embedding agnostic and allocate bits uniformly across dimensions. However, recent models produce embeddings with significant geometric structure. In this work, we investigate whether a variable bit allocation scheme can improve quantization quality under a fixed memory budget. We propose a simp

发布时间：2026-08-21 12:00

抓取时间：2026-08-21 12:32

来源机构：arXiv