Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM for Mobile: An Initial Roadmap

Daihang Chen, Yonghui Liu, Mingyi Zhou, Yanjie Zhao, Haoyu Wang, Shuai Wang, Xiao Chen, Tegawendé F. Bissyandé, Jacques Klein, Li Li

TL;DR

The paper argues for integrating LLMs into the mobile ecosystem to deliver smarter user experiences. It presents a two-phase framework (LLM Supply and LLM Use) and six urgent research directions spanning dataset preparation, mobile app development and analysis, on-device serving, security, framework APIs, and runtime monitoring. By detailing data sources, multi-modal data, on-device inference techniques, privacy-aware approaches, and intelligent runtime analysis, it outlines a path to practical mobile LLM deployment. The authors emphasize challenges such as privacy, latency, security, and the need for standardized APIs and datasets, inviting ongoing community contributions to broaden the roadmap.

Abstract

When mobile meets LLMs, mobile app users deserve to have more intelligent usage experiences. For this to happen, we argue that there is a strong need to appl LLMs for the mobile ecosystem. We therefore provide a research roadmap for guiding our fellow researchers to achieve that as a whole. In this roadmap, we sum up six directions that we believe are urgently required for research to enable native intelligence in mobile devices. In each direction, we further summarize the current research progress and the gaps that still need to be filled by our fellow researchers.

LLM for Mobile: An Initial Roadmap

TL;DR

The paper argues for integrating LLMs into the mobile ecosystem to deliver smarter user experiences. It presents a two-phase framework (LLM Supply and LLM Use) and six urgent research directions spanning dataset preparation, mobile app development and analysis, on-device serving, security, framework APIs, and runtime monitoring. By detailing data sources, multi-modal data, on-device inference techniques, privacy-aware approaches, and intelligent runtime analysis, it outlines a path to practical mobile LLM deployment. The authors emphasize challenges such as privacy, latency, security, and the need for standardized APIs and datasets, inviting ongoing community contributions to broaden the roadmap.

Abstract

When mobile meets LLMs, mobile app users deserve to have more intelligent usage experiences. For this to happen, we argue that there is a strong need to appl LLMs for the mobile ecosystem. We therefore provide a research roadmap for guiding our fellow researchers to achieve that as a whole. In this roadmap, we sum up six directions that we believe are urgently required for research to enable native intelligence in mobile devices. In each direction, we further summarize the current research progress and the gaps that still need to be filled by our fellow researchers.
Paper Structure (11 sections, 2 figures)

This paper contains 11 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Preparing Dataset for Fine-tuning LLMs
  3. SE Scenarios
  4. UX Scenarios
  5. Multi-Modal Data Processing Scenarios
  6. Applying LLMs for Mobile App Development and Analysis
  7. Serving LLM on Mobile
  8. On-device LLM Security
  9. Providing LLM-powered Framework APIs
  10. Providing LLM-powered Runtime Monitoring
  11. Conclusion

Figures (2)

  • Figure 1: Roadmap in Applying LLMs for Mobile.
  • Figure 2: The information leakage problem of on-device LLM on Android. The sensitive model representation is directly hosted on mobile devices.