Cacheback: Speculative Decoding With Nothing But Cache

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Zhiyao, Gim, In, Zhong, Lin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911289436012544
author Ma, Zhiyao
Gim, In
Zhong, Lin
author_facet Ma, Zhiyao
Gim, In
Zhong, Lin
contents We present Cacheback Decoding, a training-free and model-agnostic speculative decoding method that exploits the locality in language to accelerate Large Language Model (LLM) inference. Cacheback leverages only Least Recently Used (LRU) cache tables of token n-grams to generate draft sequences. Cacheback achieves state-of-the-art performance among comparable methods despite its minimalist design, and its simplicity allows easy integration into existing systems. Cacheback also shows potential for fast adaptation to new domains.
format Preprint
id arxiv_https___arxiv_org_abs_2511_21699
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cacheback: Speculative Decoding With Nothing But Cache
Ma, Zhiyao
Gim, In
Zhong, Lin
Computation and Language
Artificial Intelligence
We present Cacheback Decoding, a training-free and model-agnostic speculative decoding method that exploits the locality in language to accelerate Large Language Model (LLM) inference. Cacheback leverages only Least Recently Used (LRU) cache tables of token n-grams to generate draft sequences. Cacheback achieves state-of-the-art performance among comparable methods despite its minimalist design, and its simplicity allows easy integration into existing systems. Cacheback also shows potential for fast adaptation to new domains.
title Cacheback: Speculative Decoding With Nothing But Cache
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2511.21699