Vincent Tu comments

Results 30 comments of


                                            Vincent Tu

Many setup issues with OSWorld

Hi @thomas-kuntz , I'm working on #53 to help simplify the setup for the `kb` and the library. 1. Failure to import the Agent S2 classes The setup requires you...

Many setup issues with OSWorld

Hi, just following up, please check out the `osworld_setup` folder for more information. 😃

Many setup issues with OSWorld

I've updated the folder names of the knowledge base assets. This was causing the no file found error.

Evaluation Metrics, Fix-ups, Output Parsing for Evaluation

[Slack File: hotpot_dev_v1_simplified.json](https://files.slack.com/files-pri/T06G9G0703E-F07LF89C96G/hotpot_dev_v1_simplified.json)

Evaluation Metrics, Fix-ups, Output Parsing for Evaluation

Reproduction details of Agent S3 results in Paper

Hi @linyq17 , thanks for raising this issue. How are you running OSWorld? Can you double check the coding agent is being called and it's running and its generating code...

Reproduction details of Agent S3 results in Paper

Hi @linyq17 , I took a look at traj2 and compared it with a fresh run I did (35/47). The screenshot shows the IDs highlighted in red that my run...

ui tars

Hi @jsbintu , we don't support open router currently. But I'll look into implementing it. In the meantime, please try with either vllm or a Hugging Face Inference Endpoint.

Are there issues with GIMP/d16 and chrome/bb5 evaluation?

Follow up: chrome/bb5 https://github.com/xlang-ai/OSWorld/blob/51f5ddea04f72fd91d5df37bd8202c83c1daff90/evaluation_examples/examples/chrome/bb5e4c0d-f964-439c-97b6-bdb9747de3f4.json#L31-L39 I noticed for this example, the [data viewer](https://os-world.github.io/explorer.html) shows the user clicking I think it's sufficient to just click the button right above and change to...