多智能体系统Minitap在安卓任务中实现100%准确率，超越人类表现。
Do Multi-Agents Dream of Electric Screens? Achieving Perfect Accuracy on AndroidWorld Through Task Decomposition
- 分拆为六类专用智能体，避免推理干扰和重复循环
- 通过状态验证确保文本输入有效，提升执行可靠性
- 引入元认知机制自动识别死循环并切换策略
我们提出Minitap，一个在AndroidWorld基准上实现100%成功率的多智能体系统，首次完整解决全部116个任务，超越人类表现（80%）。分析表明单智能体架构失败原因包括：混合推理痕迹导致上下文污染、未被检测的静默文本输入失败、无退出机制的重复动作循环。Minitap通过三重机制应对：六类专用智能体实现认知分离；对文本输入进行确定性后验证以匹配设备状态；元认知推理识别循环并触发策略变更。消融实验显示，多智能体分解带来+21分提升，验证执行增加+7分，元认知贡献+9分。项目已开源，地址：https://github.com/minitap-ai/mobile-use。
原文摘要 · Abstract (English)
We present Minitap, a multi-agent system that achieves 100% success on the AndroidWorld benchmark, the first to fully solve all 116 tasks and surpassing human performance (80%). We first analyze why single-agent architectures fail: context pollution from mixed reasoning traces, silent text input failures undetected by the agent, and repetitive action loops without escape. Minitap addresses each failure through targeted mechanisms: cognitive separation across six specialized agents, deterministic post-validation of text input against device state, and meta-cognitive reasoning that detects cycles and triggers strategy changes. Ablations show multi-agent decomposition contributes +21 points over single-agent baselines; verified execution adds +7 points; meta-cognition adds +9 points. We release Minitap as open-source software. https://github.com/minitap-ai/mobile-use
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。