[Evaluation] Integrate BrowserGym for agent web browsing evaluation

[xingyaoww]

What problem or use case are you trying to solve?

As SWE-Bench is going to be added soon, I think it'd be good to move our evaluation codebase forward to include BrowserGym for everything web browser!

They already support:

• MiniWob++
• WebArena
• WorkArena

With more work going to be added to that benchmark soon!

cc @neubig @frankxu2004

Describe the UX of the solution you'd like

Do you have thoughts on the technical implementation?

• At least we need to figure out a way to connect OpenDevin's agent interface with BrowserGym's agent implementation as suggested by @frankxu2004

Describe alternatives you've considered

Additional context

[xingyaoww]

Re-open since the evaluation is not added yet.

[neubig]

We have a separate issue for evaluation #1471 , so maybe it can be closed?

[xingyaoww]

ohhh good point!