Research

Paper

AI LLM March 17, 2026

VQKV: High-Fidelity and High-Ratio Cache Compression via Vector-Quantization

Authors

Yixuan Wang, Qingyu Shi, Jiayu Zhou, Dianbo Liu, Ziwei He, Zhouhan Lin

Abstract

The growing context length of Large Language Models (LLMs) enlarges the Key-Value (KV) cache, limiting deployment in resource-limited environments. Prior training-free approaches for KV cache compression typically rely on low-rank approximation or scalar quantization, which fail to simultaneously achieve high compression ratios and high reconstruction fidelity. We propose VQKV, a novel, training-free method introducing vector quantization (VQ) to obtain highly compressed KV representations while preserving high model fidelity, allowing for the representation of thousands of floating-point values with just a few integer indices. As a result, VQKV achieves an 82.8\% compression ratio on LLaMA3.1-8B while retaining 98.6\% of the baseline performance on LongBench and enabling 4.3x longer generation length on the same memory footprint.

Metadata

arXiv ID: 2603.16435
Provider: ARXIV
Primary Category: cs.CL
Published: 2026-03-17
Fetched: 2026-03-18 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.16435v1</id>\n    <title>VQKV: High-Fidelity and High-Ratio Cache Compression via Vector-Quantization</title>\n    <updated>2026-03-17T12:16:06Z</updated>\n    <link href='https://arxiv.org/abs/2603.16435v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.16435v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>The growing context length of Large Language Models (LLMs) enlarges the Key-Value (KV) cache, limiting deployment in resource-limited environments. Prior training-free approaches for KV cache compression typically rely on low-rank approximation or scalar quantization, which fail to simultaneously achieve high compression ratios and high reconstruction fidelity. We propose VQKV, a novel, training-free method introducing vector quantization (VQ) to obtain highly compressed KV representations while preserving high model fidelity, allowing for the representation of thousands of floating-point values with just a few integer indices. As a result, VQKV achieves an 82.8\\% compression ratio on LLaMA3.1-8B while retaining 98.6\\% of the baseline performance on LongBench and enabling 4.3x longer generation length on the same memory footprint.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-17T12:16:06Z</published>\n    <arxiv:primary_category term='cs.CL'/>\n    <author>\n      <name>Yixuan Wang</name>\n    </author>\n    <author>\n      <name>Qingyu Shi</name>\n    </author>\n    <author>\n      <name>Jiayu Zhou</name>\n    </author>\n    <author>\n      <name>Dianbo Liu</name>\n    </author>\n    <author>\n      <name>Ziwei He</name>\n    </author>\n    <author>\n      <name>Zhouhan Lin</name>\n    </author>\n  </entry>"
}