jacob-valdez/synthux-economy-r4-w018
收藏资源简介:
SynthUX计算机使用数据集(经济模拟r4)是一个基于模拟环境的计算机使用轨迹数据集,由SynthUX生成。每个记录代表一个工作人员在模拟公司设备会话中的行为:目标扩展为节点树,终端节点通过低级别鼠标/键盘输入驱动真实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),观察到的轨迹以屏幕视频和每个节点的帧形式记录。数据集特点包括应用原生多应用程序执行(无工作台覆盖)、每个参与者内容多样性、跨操作系统消息传递以及大型多公司经济模拟。每个经济模拟作为独立的数据集仓库发布。内容包含data.jsonl文件(每个渲染轨迹一个记录)、screenshots文件夹(每个终端节点的帧,在节点的最后一个输入事件时拍摄)和videos文件夹(完整屏幕录制视频)。记录模式包括轨迹ID、参与者ID、公司、角色ID、标题、环境(模拟器类型)、自然语言目标、节点树、密集低级别输入事件、因果模拟器状态观察、对齐信息以及媒体文件(视频和帧)。统计信息显示本构建包含36个轨迹、843帧、6个公司,环境分布为browser-os(9个)、macos-web-next(9个)和windows-web-next(18个)。数据集使用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 r4 is a dataset of grounded visual computer-use trajectories generated by SynthUX. Each record represents one workers device session inside a simulated company: a goal expands into a node tree, terminal nodes drive real desktop-simulator apps (such as Terminal, Notes, VS Code, Browser, Slack/Teams, Mail, Sheets, Slides, etc.) through low-level mouse/keyboard input, and the observed trajectory is recorded as a screen video plus per-node frames. Key features include app-native multi-application execution (no 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/` (a frame per terminal node, taken at that nodes last input event), and `videos/` (the full screen recording). The record schema covers fields like trajectory_id, actor_id, company, role_id, title, env (simulator type), goal, tree.nodes, trajectory.input_events, trajectory.observations, trajectory.alignment, and media files (video and frames). Statistics for this build show 36 trajectories, 843 frames, 6 companies, with environment distribution: browser-os (9), macos-web-next (9), windows-web-next (18). Provenance indicates generation with SynthUX, where observations are causal (mutation channel = low_level_input) and no high-level app mutation is used during capture.




