Tianbao Xie

Depends on the scale you want, if scale is large, try rebuild the image on the cloud service you are using following https://github.com/xlang-ai/OSWorld/blob/main/desktop_env/server/README.md , and follow their reset API.

Hi @cjfcsjt Thank you for raising this question! The following script is my `qwen25vl_agent.py` implementation for OSWorld. You can place this file under the `OSWorld/mm_agents` directory in your project. ---...

For your question 2, when benchmarking OSWorld from prompting qwen2.5-vl, the thought is included in the response, for example: ``` Action: Click on the 'Relaunch to update' button to proceed...

Hi, Thanks for the question. Yes, the a11y tree API sometimes cannot offer the full tree as we expected, which is a matter of ecosystem and community support. That's one...

We actually add special handling for this issue... But if you met this problem, maybe check out https://github.com/xlang-ai/OSWorld/issues/182 would be helpful.

Hi, sorry we don't have the stepped screenshots yet.

Hi, Actually, we have a self-hosted version of macOS, but due to the potential license issues, we cannot make that public. Best, Tianbao

Hi, thanks for the question, have you fixed it already? I can provide further help if you like.

Hi, Thanks for the question. In the long run, I think the screenshot setting is the most promising one, and to be more aggressive, actually, we can let the video...

The OSWorld environment, when paired with Docker, indeed relies on the core system supporting KVM (Kernel-based Virtual Machine) for optimal performance, as KVM provides hardware-assisted virtualization. This dependency traditionally limits...