Blog
LLMs & Texto
Ablation, Statistical Inference, and Validation for KV-Cache Compression
arXiv:2607.09683v1 Announce Type: new Abstract: 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 effe...
arXiv cs.LG
·Paolo D'Alberto, Ashish Siarasao, Elliott Delaye, Rajeev Patwari
·