LLM Cache Bandit Revisited: Addressing Query Heterogeneity for Cost-Effective LLM Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Hantao, Xie, Hong, Lian, Defu, Chen, Enhong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916957411868672
author Yang, Hantao
Xie, Hong
Lian, Defu
Chen, Enhong
author_facet Yang, Hantao
Xie, Hong
Lian, Defu
Chen, Enhong
contents This paper revisits the LLM cache bandit problem, with a special focus on addressing the query heterogeneity for cost-effective LLM inference. Previous works often assume uniform query sizes. Heterogeneous query sizes introduce a combinatorial structure for cache selection, making the cache replacement process more computationally and statistically challenging. We treat optimal cache selection as a knapsack problem and employ an accumulation-based strategy to effectively balance computational overhead and cache updates. In theoretical analysis, we prove that the regret of our algorithm achieves an $O(\sqrt{MNT})$ bound, improving the coefficient of $\sqrt{MN}$ compared to the $O(MN\sqrt{T})$ result in Berkeley, where $N$ is the total number of queries and $M$ is the cache size. Additionally, we also provide a problem-dependent bound, which was absent in previous works. The experiment rely on real-world data show that our algorithm reduces the total cost by approximately 12\%.
format Preprint
id arxiv_https___arxiv_org_abs_2509_15515
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LLM Cache Bandit Revisited: Addressing Query Heterogeneity for Cost-Effective LLM Inference
Yang, Hantao
Xie, Hong
Lian, Defu
Chen, Enhong
Computation and Language
This paper revisits the LLM cache bandit problem, with a special focus on addressing the query heterogeneity for cost-effective LLM inference. Previous works often assume uniform query sizes. Heterogeneous query sizes introduce a combinatorial structure for cache selection, making the cache replacement process more computationally and statistically challenging. We treat optimal cache selection as a knapsack problem and employ an accumulation-based strategy to effectively balance computational overhead and cache updates. In theoretical analysis, we prove that the regret of our algorithm achieves an $O(\sqrt{MNT})$ bound, improving the coefficient of $\sqrt{MN}$ compared to the $O(MN\sqrt{T})$ result in Berkeley, where $N$ is the total number of queries and $M$ is the cache size. Additionally, we also provide a problem-dependent bound, which was absent in previous works. The experiment rely on real-world data show that our algorithm reduces the total cost by approximately 12\%.
title LLM Cache Bandit Revisited: Addressing Query Heterogeneity for Cost-Effective LLM Inference
topic Computation and Language
url https://arxiv.org/abs/2509.15515