# GSRQ: Gain-Shape Residual Quantization for Sub-1-bit KV Cache

Preprint · Soosung Kim, Minjae Park, Eui-Young Chung, Jaeyong Chung

- Archive id: arxiv:2607.01065
- Published: 2026-07-01
- arXiv: https://arxiv.org/abs/2607.01065
- Citations: 0
- Source URL: https://arxiv.org/abs/2607.01065
- PDF: https://arxiv.org/pdf/2607.01065
- Provenance: arxiv

## Abstract

The deployment of Large Language Models (LLMs) with extended context windows is increasingly constrained by the linear growth of Key-Value (KV) cache memory. Vector Quantization (VQ), particularly Residual Quantization (RQ), is a promising approach for pushing KV cache storage toward the sub-1-bit regime by progressively encoding residuals with small codebooks. However, most VQ methods still rely on standard $\\ell\_2$ $K$-means as the core codebook-learning primitive. We identify a subtle high-dimensional issue of this primitive: Euclidean centroid averaging can induce centroid shrinkage, which weakens the angular alignment term in the $\\ell\_2$ distortion and makes directional preservation harder. To address this issue, we propose Gain-Shape $K$-means (GSKM), a drop-in replacement for $K$-means that improves directional fidelity while matching, and in some regimes improving, $\\ell\_2$ distortion. We then build Gain-Shape Residual Quantization (GSRQ) by incorporating a weighted extension of GSKM into an RQ pipeline. On LLaMA-3-8B, GSRQ substantially improves over strong KV cache quantization baselines across bit rates. At 1-bit, it improves the average accuracy across LongBench tasks from 11.34 to 33.54, a gain of 22.20 percentage points over VQLLM.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

Topics: [Efficient Training & Inference](https://space-kitty.com/topics/efficient-ai)

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.