发表机构
Advanced Micro Devices, Inc.（超威半导体公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究对Turbo-Quant和SpectralQuant KV缓存压缩进行系统比较，用统计验证方法评估非支配方案，发现基于特征基的方法在重尾数据上因协方差不稳定失败，在结构化模式下表现出色，有效语义维度适应校准预算。
AI 中文摘要
本研究系统地比较了Turbo-Quant和SpectralQuant KV缓存压缩，通过一种统计验证方法评估非支配方案，包括带Beta Lloyd-Max和QJL的WHT旋转，该方法将编解码器系统差异与实现差异分开。关键发现表明，基于特征基的方法因协方差不稳定在重尾数据上失败，但在结构化模式下表现出色，有效语义维度（$d_{eff}$）适应校准预算而非真实数据秩。
英文摘要
This study systematically compares Turbo-Quant and SpectralQuant KV-cache compression, evaluating non-dominated schemes, including WHT rotation with Beta Lloyd-Max and QJL, through a statistical validation methodology that separates systematic codec differences from implementation variance. Key findings reveal that while eigenbasis-based methods fail on heavy-tailed data due to covariance instability, they excel in structured regimes, with the effective semantic dimension ($d_{eff}$) adapting to calibration budgets rather than true data rank. (this is an abstract of the abstract thank you )
Comments15 pages, 8 figures, minimum number of citations
