Saved in:
Bibliographic Details
Main Authors: Wang, Tuowei, Li, Fengzu, Sun, Yanfan, Gao, Wei, Ren, Ju
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.16786
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911690878091264
author Wang, Tuowei
Li, Fengzu
Sun, Yanfan
Gao, Wei
Ren, Ju
author_facet Wang, Tuowei
Li, Fengzu
Sun, Yanfan
Gao, Wei
Ren, Ju
contents Large language models (LLMs) are increasingly needed for interactive mobile applications, but high-quality models exceed the limited DRAM available on smartphones. Flash storage can hold larger models, yet flash-backed inference is slow because autoregressive decoding repeatedly invokes the target model and incurs costly I/O. We observe that speculative decoding is a natural fit for this setting: a small draft model can remain in DRAM, while a larger flash-resident target model verifies multiple candidate tokens per invocation. However, existing methods assume server-class accelerators and fail to account for prolonged I/O latency, limited computation parallelism, and irregular speculation execution. We present Lever, an end-to-end system for efficient flash-backed LLM inference on smartphones. Lever jointly optimizes the three stages of speculative decoding under mobile constraints. For drafting, it builds token trees using an I/O- and compute-aware gain-cost objective. For verification, it prunes low-value branches through early-exit prediction to reduce target-model computation. For execution, it maps speculation efficiently across mobile CPU-NPU hardware to improve utilization. Comprehensive evaluations show that Lever reduces inference latency by an average of 2.93x over baseline flash-offloaded inference and 1.50x over conventional speculative decoding, narrowing the latency gap between flash-backed and memory-resident LLM inference.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16786
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Lever: Speculative LLM Inference on Smartphones
Wang, Tuowei
Li, Fengzu
Sun, Yanfan
Gao, Wei
Ren, Ju
Machine Learning
Large language models (LLMs) are increasingly needed for interactive mobile applications, but high-quality models exceed the limited DRAM available on smartphones. Flash storage can hold larger models, yet flash-backed inference is slow because autoregressive decoding repeatedly invokes the target model and incurs costly I/O. We observe that speculative decoding is a natural fit for this setting: a small draft model can remain in DRAM, while a larger flash-resident target model verifies multiple candidate tokens per invocation. However, existing methods assume server-class accelerators and fail to account for prolonged I/O latency, limited computation parallelism, and irregular speculation execution. We present Lever, an end-to-end system for efficient flash-backed LLM inference on smartphones. Lever jointly optimizes the three stages of speculative decoding under mobile constraints. For drafting, it builds token trees using an I/O- and compute-aware gain-cost objective. For verification, it prunes low-value branches through early-exit prediction to reduce target-model computation. For execution, it maps speculation efficiently across mobile CPU-NPU hardware to improve utilization. Comprehensive evaluations show that Lever reduces inference latency by an average of 2.93x over baseline flash-offloaded inference and 1.50x over conventional speculative decoding, narrowing the latency gap between flash-backed and memory-resident LLM inference.
title Lever: Speculative LLM Inference on Smartphones
topic Machine Learning
url https://arxiv.org/abs/2605.16786