arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.22009cs.AI
收藏

OpenPhone: 移动代理基础模型

OpenPhone: Mobile Agentic Foundation Models

The University of Hong Kong(香港大学)

Yangqin Jiang, Chao Huang

AI总结:

OpenPhone通过设备-云协作提升移动GUI代理性能,结合设备端高效模型与云端强大能力,实现成本降低与性能提升的平衡。

摘要:

随着多模态大语言模型(MLLMs)的发展,构建GUI代理系统已成为一个日益有前途的方向--尤其是对于移动平台,鉴于其丰富的应用生态系统和直观的触控交互。然而,移动GUI代理面临一个关键难题:真正在设备上的模型(4B或更小)性能不足,而有能力的模型(从7B开始)要么太大无法在移动平台上部署,要么成本过高(例如,仅云上的封闭源MLLMs)。为了解决这一问题,我们提出了OpenPhone,一种移动GUI代理系统,通过设备-云协作来利用在设备上的模型的成本效率和云模型的高能力,同时避免其缺点。具体而言,OpenPhone通过在合成GUI数据上进行两阶段SFT->GRPO训练来增强Qwen2.5-VL-3B,整合一个高效的长推理和内存管理机制以在资源紧张的情况下利用历史交互,并默认在设备上执行--仅将具有挑战性的子任务通过实时复杂度评估提升到云上。在在线AndroidLab基准和多样应用上的实验表明,OpenPhone与更大模型相匹配或接近,同时显著降低了云成本。

英文摘要:

With the advancement of multimodal large language models (MLLMs), building GUI agent systems has become an increasingly promising direction--especially for mobile platforms, given their rich app ecosystems and intuitive touch interactions. Yet mobile GUI agents face a critical dilemma: truly on-device models (4B or smaller) lack sufficient performance, while capable models (starting from 7B) are either too large for mobile deployment or prohibitively costly (e.g., cloud-only closed-source MLLMs). To resolve this, we propose OpenPhone, a mobile GUI agent system that leverages device-cloud collaboration to tap the cost-efficiency of on device models and the high capability of cloud models, while avoiding their drawbacks. Specifically, OpenPhone enhances Qwen2.5-VL-3B via two-stage SFT->GRPO training on synthetic GUI data for strong decision-making, integrates an efficient long-reasoning and memory management mechanism to utilize historical interactions under tight resources, and defaults to on-device execution--only escalating challenging subtasks to the cloud via real-time complexity assessment. Experiments on the online AndroidLab benchmark and diverse apps show OpenPhone matches or nears larger models, with a significant reduction in cloud costs.

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。

↑