FAEDKV: Infinite-Window Fourier Transform for Unbiased KV Cache Compression

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Runchao, Fu, Yao, Sheng, Mu, Long, Xianxuan, Yu, Haotian, Li, Pan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912504135811072
author Li, Runchao
Fu, Yao
Sheng, Mu
Long, Xianxuan
Yu, Haotian
Li, Pan
author_facet Li, Runchao
Fu, Yao
Sheng, Mu
Long, Xianxuan
Yu, Haotian
Li, Pan
contents The efficacy of Large Language Models (LLMs) in long-context tasks is often hampered by the substantial memory footprint and computational demands of the Key-Value (KV) cache. Current compression strategies, including token eviction and learned projections, frequently lead to biased representations -- either by overemphasizing recent/high-attention tokens or by repeatedly degrading information from earlier context -- and may require costly model retraining. We present FAEDKV (Frequency-Adaptive Infinite-Window for KV cache), a novel, training-free KV cache compression framework that ensures unbiased information retention. FAEDKV operates by transforming the KV cache into the frequency domain using a proposed Infinite-Window Fourier Transform (IWDFT). This approach allows for the equalized contribution of all tokens to the compressed representation, effectively preserving both early and recent contextual information. A preliminary frequency ablation study identifies critical spectral components for layer-wise, targeted compression. Experiments on LongBench benchmark demonstrate FAEDKV's superiority over existing methods by up to 22\%. In addition, our method shows superior, position-agnostic retrieval accuracy on the Needle-In-A-Haystack task compared to compression based approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2507_20030
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FAEDKV: Infinite-Window Fourier Transform for Unbiased KV Cache Compression
Li, Runchao
Fu, Yao
Sheng, Mu
Long, Xianxuan
Yu, Haotian
Li, Pan
Computation and Language
The efficacy of Large Language Models (LLMs) in long-context tasks is often hampered by the substantial memory footprint and computational demands of the Key-Value (KV) cache. Current compression strategies, including token eviction and learned projections, frequently lead to biased representations -- either by overemphasizing recent/high-attention tokens or by repeatedly degrading information from earlier context -- and may require costly model retraining. We present FAEDKV (Frequency-Adaptive Infinite-Window for KV cache), a novel, training-free KV cache compression framework that ensures unbiased information retention. FAEDKV operates by transforming the KV cache into the frequency domain using a proposed Infinite-Window Fourier Transform (IWDFT). This approach allows for the equalized contribution of all tokens to the compressed representation, effectively preserving both early and recent contextual information. A preliminary frequency ablation study identifies critical spectral components for layer-wise, targeted compression. Experiments on LongBench benchmark demonstrate FAEDKV's superiority over existing methods by up to 22\%. In addition, our method shows superior, position-agnostic retrieval accuracy on the Needle-In-A-Haystack task compared to compression based approaches.
title FAEDKV: Infinite-Window Fourier Transform for Unbiased KV Cache Compression
topic Computation and Language
url https://arxiv.org/abs/2507.20030