arXivDaily arXiv每日学术速递 周一至周五更新
​通知:联邦节假日调休:哥伦布日 / 原住民日(Columbus Day / Indigenous Peoples' Day),今日(13日)无更新。
arXiv 2604.14159cs.CLcs.AI

HUOZIIME:一种基于大语言模型的设备端输入法用于深度个性化

HUOZIIME: An On-Device LLM-enhanced Input Method for Deep Personalization

Harbin Institute of Technology(哈尔滨理工大学)

Baocai Shan, Yuzhuang Xu, Wanxiang Che

AI总结:

本文提出HUOZIIME,一种基于大语言模型的设备端输入法,通过合成个性化数据训练基础模型,结合分层记忆机制实现高效个性化文本生成,实验表明其在设备端执行高效且个性化质量高。

摘要:

移动输入法编辑器(IMEs)是文本输入的主要接口,但受限于手动输入,难以生成个性化文本。尽管轻量级大语言模型(LLMs)使设备端辅助生成成为可能,但实现深度个性化、隐私保护和实时生成的IME仍面临根本挑战。为此,我们提出了HUOZIIME,一种由LLM驱动的个性化设备端IME。我们通过在合成个性化数据上微调基础LLM,赋予HUOZIIME初始的人类样预测能力。值得注意的是,设计了分层记忆机制以持续捕获和利用用户特定的输入历史。此外,我们进行了针对设备端LLM基于IME部署的系统优化,确保在移动约束下实现高效且响应迅速的操作。实验表明,HUOZIIME在设备端执行高效且具有高保真度的记忆驱动个性化。代码和包可在https://github.com/Shan-HIT/HuoziIME获取。

英文摘要:

Mobile input method editors (IMEs) are the primary interface for text input, yet they remain constrained to manual typing and struggle to produce personalized text. While lightweight large language models (LLMs) make on-device auxiliary generation feasible, enabling deeply personalized, privacy-preserving, and real-time generative IMEs poses fundamental challenges.To this end, we present HUOZIIME, a personalized on-device IME powered by LLM. We endow HUOZIIME with initial human-like prediction ability by post-training a base LLM on synthesized personalization data. Notably, a hierarchical memory mechanism is designed to continually capture and leverage user-specific input history. Furthermore, we perform systemic optimizations tailored to on-device LLMbased IME deployment, ensuring efficient and responsive operation under mobile constraints.Experiments demonstrate efficient on-device execution and high-fidelity memory-driven personalization. Code and package are available at https://github.com/Shan-HIT/HuoziIME.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑