RecreationWorld：可扩展可验证的混合计算机使用智能体环境
原标题：Paper page - RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents
AI 摘要
研究者提出 RecreationWorld，一个面向混合计算机使用智能体（CUA）的五平台框架，覆盖 Ubuntu、macOS、Windows、Android 和 Web，通过让智能体复现运行中的参考应用来提供可验证的执行奖励。基于高质量开源应用扩展轨迹生成后，模型在五个分布外编码与混合计算机使用基准上均有提升，并更频繁地验证渲染输出。团队同时发布 RecreationBench，包含 250 个跨领域跨平台任务，GPT-6 Astra 总体得分 58.1%，但仅 2.8% 任务通过全部程序化测试，显示智能体在复现交互与计算结果方面仍弱于静态界面结构。
正文节选
RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents Abstract Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development through code and the command line. Real digital work requires both, interleaved rather than stacked end to end. We study hybrid CUAs that autonomously decide when to explore an interface, implement software, and run and visually verify their artifacts. We introduce RecreationWorld, a five-pl