Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents
作者: Tingyu Qu, Weigao Sun, Yuecheng Liu, Yucheng Zhao, Yi Zhu, Yifeng Ding, Qiyi Wang, Sihan Cao, Pengkun Jiao, Hanlei Xie, Xiongwei Wu, Qichao Wang, Haodong Zhang, Jiajun Liu, Yuhao Wang, Yuqing Xie, Junpeng Zhao, Long Chen, Ming Ma, Sihan Yang, Ziwang Zhao, Yanhao Jia, Liangquan Gong, Feida Zhu, Yiran Zhong, Steven Hoi
分类: cs.AI
发布日期: 2026-09-24
备注: https://tongyi-mai.github.io/Qwen-Planner-Agent/
💡 一句话要点
提出Qwen-Planner-Agent以解决移动规划代理的开发与迭代问题
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 移动规划 人工智能 闭环框架 模型训练 强化学习 数据生成 代理系统 性能优化
📋 核心要点
- 现有方法在复杂、长时间的移动规划任务中面临代理可靠性不足和真实设备交互成本高的问题。
- 论文提出的Qwen-Planner-Agent通过闭环AI-for-AI框架实现数据生成、模型训练和部署的高效连接。
- 实验结果表明,Qwen-Planner-Agent在MobilePA-Bench上表现最佳,相较于基线模型在工具使用、记忆、技能和子代理协调方面都有显著提升。
📝 摘要(中文)
大型语言模型的快速发展使得人工智能从被动内容生成转向主动参与工程和科学发现的工作流程。本文探讨了人工智能是否可以作为开发对象,同时也是构建下一代人工智能系统的积极参与者。我们在一个闭环的AI-for-AI框架内构建了Qwen-Planner-Agent,以实现可扩展的开发和迭代改进。该框架通过共享的行动-反馈-验证合同连接数据生产、模型训练和部署,展示了在复杂的移动规划任务中如何提高代理的可靠性和开发的可扩展性。
🔬 方法详解
问题定义:本论文旨在解决在复杂移动规划任务中,现有代理系统在可靠性和开发可扩展性方面的不足。现有方法在真实设备交互中成本高,难以进行有效的迭代改进。
核心思路:Qwen-Planner-Agent通过构建一个闭环的AI-for-AI框架,连接数据生成、模型训练和部署,利用专门的代理构建任务并收集交互轨迹,从而实现高效的数据生成和模型优化。
技术框架:该框架包含三个主要模块:AI for Data、AI for Training和AI驱动的模型-工具协同进化。AI for Data负责数据生成和反馈,AI for Training结合监督学习和强化学习进行模型训练,而AI模块则通过执行证据驱动的循环来优化模型和工具的协同。
关键创新:引入了Competence-Aware Reward-and-Advantage Engineering (CARE)方法,旨在降低推理和工具使用的成本,同时保持任务性能。这一创新使得模型在复杂任务中的表现显著提升。
关键设计:在模型训练中,采用了混合环境的在线强化学习策略,并通过人类引导的代理数据飞轮来平衡训练数据。此外,设计了结构化的行动反馈机制,以便于模型和工具的协调适应。
🖼️ 关键图片
📊 实验亮点
在MobilePA-Bench的评估中,Qwen-Planner-Agent在所有模型和系统中表现最佳,较基线模型在工具使用、记忆、技能和子代理协调方面均有显著提升,展示了其在复杂任务中的优越性能。
🎯 应用场景
Qwen-Planner-Agent的研究成果在移动机器人、自动驾驶、智能制造等领域具有广泛的应用潜力。通过提高代理的可靠性和开发效率,该框架能够加速智能系统的迭代与优化,推动相关技术的进步与普及。
📄 摘要(原文)
The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI systems? We explore this question by building Qwen-Planner-Agent within a closed-loop AI-for-AI framework for scalable development and iterative improvement. Mobile planning offers a demanding test of this approach: complex, long-horizon tasks challenge agent reliability, while costly real-device interaction limits development scalability. The framework connects data production, model training, and deployment through a shared action-feedback-verification contract. (i) AI for Data builds a human-gated agentic data flywheel in which specialized agents construct tasks, collect interaction trajectories, curate and balance training data, and use training feedback to guide subsequent data generation. (ii) AI for Training combines a supervised planning cold start with hybrid-environment online agentic reinforcement learning, where we introduce Competence-Aware Reward-and-Advantage Engineering (CARE) to reduce reasoning and tool-use costs while preserving task performance. (iii) AI drives model--harness co-evolution through an execution-evidence-driven loop that orchestrates memory, skills, and tools at runtime and feeds structured action feedback and preserved failure traces back into coordinated model and harness adaptation. Qwen-Planner-Agent achieves the best overall performance among all evaluated models and systems on MobilePA-Bench, improving over its base model across tool use, memory, skills, and sub-agent coordination. Further evaluations of our model show improvements across non-mobile agentic benchmarks while largely preserving general capabilities.