OSWorld Model Card
这是可索引项目报告证据页：它保留 OSWorld Model Card 的源材料入口、机制线索和限制提醒；正文仍需 reader/editor 与 academic public-copy review 后才能当作最终结论引用。
OSWorld Model Card
| Field | Value |
|---|---|
| Repository | xlang-ai/OSWorld |
| Category | Computer-use Agent OS Benchmark |
| Stars / forks snapshot | 2900 / 463 |
| Language | Python |
| License | Apache-2.0 |
| Raw capture | raw-github/xlang-ai_osworld.md |
| Updated by | hourly public metadata update, 2026-05-24 |
1. Role in Self Evolve
OSWorld 是 NeurIPS 2024 的真实计算机环境 multimodal agent benchmark，提供桌面环境、数据、运行脚本和 OSWorld-Verified 更新，用于衡量 CUA/GUI agent 能力。
2. Working Principle
真实桌面任务 -> 虚拟机/云环境 -> multimodal agent actions -> benchmark result comparison
3. Evidence Path
web GitHub page observed README, 1,444 commits, Apache-2.0 license, 2.9k stars, 463 forks, project page/paper/data links, and OSWorld-Verified update signals; shell GitHub API was blocked by DNS and gh auth was invalid, so this run marks freshness as web-page observed rather than API verified.
4. Teaching Use
Use this card to explain whether a repository is a runtime, benchmark, harness-evolution loop, memory/skill substrate, or resource index. The key reading path is: raw capture -> classification row -> public site card -> project report -> aggregate GitHub analysis.
5. Limits
当前未克隆源码，未运行 benchmark 或 SDK examples；star/fork/commit 快照来自公开 GitHub 页面文本或可见页面片段。