Replies: 1 comment
-
|
Hi @xhluca , I don't think you will be able to see that post-hoc in the traces. The easiest way to check would be to set yourself in debug mode, place a few breakpoints in the code and inspect the self.config variable. You could also go down to webarena's evaluator function and try to understand at which point the URLs are checked. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
This code is used to replace the pattern
__SHOPPING__into the correct URL. Is there a way to inspect the replacement post-hoc (after inference) to ensure that tasks were correctly inspected?Here's the replacement code:
BrowserGym/browsergym/webarena/src/browsergym/webarena/task.py
Lines 56 to 63 in 66cf402
I have noticed that in many cases a reward of 0 is assigned even though the correct URL was navigated at certain point
Beta Was this translation helpful? Give feedback.
All reactions