SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Liu, Hongyao, Zhai, Liuqun, Wang, Junyi, Fang, Zhengru
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909012993245184
author Liu, Hongyao
Zhai, Liuqun
Wang, Junyi
Fang, Zhengru
author_facet Liu, Hongyao
Zhai, Liuqun
Wang, Junyi
Fang, Zhengru
contents Efficient inference for on-device Large Language Models (LLMs) remains challenging due to limited hardware resources and the high cost of the prefill stage, which processes the full input context to construct Key-Value (KV) caches. We present SparKV, an adaptive KV loading framework that combines cloud-based KV streaming with on-device computation. SparKV models the cost of individual KV chunks and decides whether each chunk should be streamed or computed locally, while overlapping the two execution paths to reduce latency. To handle fluctuations in wireless connectivity and edge resource availability, SparKV further refines offline-generated schedules at runtime to rebalance communication and computation costs. Experiments across diverse datasets, LLMs, and edge devices show that SparKV reduces Time-to-First-Token by 1.3$x-5.1x with negligible impact on response quality, while lowering per-request energy consumption by 1.5x to 3.3x, demonstrating its robustness and practicality for real-world on-device deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2604_21231
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference
Liu, Hongyao
Zhai, Liuqun
Wang, Junyi
Fang, Zhengru
Networking and Internet Architecture
Artificial Intelligence
Performance
Efficient inference for on-device Large Language Models (LLMs) remains challenging due to limited hardware resources and the high cost of the prefill stage, which processes the full input context to construct Key-Value (KV) caches. We present SparKV, an adaptive KV loading framework that combines cloud-based KV streaming with on-device computation. SparKV models the cost of individual KV chunks and decides whether each chunk should be streamed or computed locally, while overlapping the two execution paths to reduce latency. To handle fluctuations in wireless connectivity and edge resource availability, SparKV further refines offline-generated schedules at runtime to rebalance communication and computation costs. Experiments across diverse datasets, LLMs, and edge devices show that SparKV reduces Time-to-First-Token by 1.3$x-5.1x with negligible impact on response quality, while lowering per-request energy consumption by 1.5x to 3.3x, demonstrating its robustness and practicality for real-world on-device deployment.
title SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference
topic Networking and Internet Architecture
Artificial Intelligence
Performance
url https://arxiv.org/abs/2604.21231