jacob-valdez/synthux-economy-r5-w089
收藏资源简介:
SynthUX计算机使用数据集是一个基于模拟环境的计算机使用轨迹数据集,专门用于经济模拟(版本r5)。该数据集包含由SynthUX生成的基于视觉的计算机使用轨迹,每个记录代表一个模拟公司内工作人员的设备会话。具体来说,目标会扩展为节点树,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),具有每个参与者的内容多样性、跨操作系统消息传递以及大型多公司经济模拟。每个经济模拟都作为独立的数据集存储库发布。数据内容包括:data.jsonl(每个渲染轨迹的记录)、screenshots/<trajectory>/<node_id>.png(每个终端节点的帧,在节点的最后一个输入事件时捕获)和videos/<trajectory>.webm(完整屏幕录制)。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、输入事件、观察结果、对齐信息和媒体文件。统计信息显示,该构建包含10个轨迹、10个视频、215帧、5个公司,所有环境均为browser-os。数据来源为SynthUX生成,观察结果是因果的,使用低级别输入通道,未使用高级应用程序突变。
The SynthUX Computer Usage Dataset is a simulated-environment-based computer usage trajectory dataset specifically designed for economic simulations (version r5). This dataset contains vision-based computer usage trajectories generated by SynthUX, where each record represents a device session of a worker within a simulated company. Specifically, goals are expanded into node trees, and terminal nodes drive real desktop emulator applications (e.g., terminal, Notes, VS Code, browser, Slack/Teams, Mail, Excel, PowerPoint, etc.) via low-level mouse/keyboard inputs. The observed trajectories are recorded as screen videos and frames for each node. The dataset supports native multi-application execution (no workspace overlay), featuring content diversity per participant, cross-OS messaging, and large-scale multi-company economic simulations. Each economic simulation is released as an independent dataset repository. The dataset contents include: data.jsonl (records for each rendered trajectory), screenshots/<trajectory>/<node_id>.png (frames for each terminal node, captured at the last input event of the node), and videos/<trajectory>.webm (full screen recordings). The record schema includes trajectory ID, participant ID, company, role ID, title, environment (emulator type), natural language goal, node tree, input events, observations, alignment information, and media files. Statistical information shows that this build contains 10 trajectories, 10 videos, 215 frames, and 5 companies, with all environments being browser-os. The dataset is generated by SynthUX, where observations are causal, using low-level input channels without relying on high-level application mutations.




