jacob-valdez/synthux-economy-r2-w010
收藏资源简介:
SynthUX计算机使用数据集(经济模拟r2)是一个合成数据集,用于计算机使用(GUI代理)任务。它包含由SynthUX生成的基础视觉计算机使用轨迹,每个记录代表一个工作者在模拟公司内的设备会话。在会话中,自然语言目标扩展为节点树,终端节点通过低级别鼠标和键盘输入驱动真实的桌面模拟器应用程序(如终端、笔记、VS Code、浏览器、Slack/Teams、邮件、表格、幻灯片等),观察到的轨迹被记录为屏幕视频和每个节点的帧。数据集支持应用原生多应用程序执行(无工作台覆盖),具有每个参与者内容的多样性、跨操作系统消息传递和大型多公司经济环境。每个经济模拟都作为独立的数据集仓库发布。数据集内容包括:data.jsonl(每个渲染轨迹一条记录)、screenshots/(每个轨迹和节点的PNG帧,在节点最后一个输入事件处捕获)和videos/(完整屏幕录制的webm文件)。记录模式详细描述了轨迹ID、参与者ID、公司、角色、目标、节点树、输入事件、观察、对齐和媒体引用等字段。本构建包含36个轨迹、903帧和6家公司,按环境分布为browser-os(12个轨迹)、macos-web-next(17个轨迹)和windows-web-next(7个轨迹)。数据集由SynthUX生成,观察是因果的(突变通道为低级别输入),捕获过程中未使用高级别应用程序突变。
The SynthUX Computer-Use Dataset (economy sim r2) is a synthetic dataset for computer-use (GUI agents) tasks. It contains grounded visual computer-use trajectories generated by SynthUX, where each record represents one workers device session inside a simulated company. In a session, a natural-language goal expands into a node tree, terminal nodes drive real desktop-simulator applications (e.g., Terminal, Notes, VS Code, Browser, Slack/Teams, Mail, Sheets, Slides) through low-level mouse and keyboard input, and the observed trajectory is recorded as a screen video plus per-node frames. The dataset features app-native multi-application execution (without a workbench overlay), per-actor content diversity, cross-OS messaging, and large multi-company economies. Each economy simulation is published as its own dataset repository. Contents include: data.jsonl (one record per rendered trajectory), screenshots/ (PNG frames per trajectory and node, taken at the nodes last input event), and videos/ (full screen recordings in webm format). The record schema details fields such as trajectory_id, actor_id, company, goal, tree.nodes, trajectory.input_events, trajectory.observations, trajectory.alignment, and media references. This build includes 36 trajectories, 903 frames, and 6 companies, distributed by environment as browser-os (12 trajectories), macos-web-next (17 trajectories), and windows-web-next (7 trajectories). The dataset is generated with SynthUX, with causal observations (mutation channel = low_level_input) and no high-level app mutation used during capture.




