[Question] Have you ever considered to use a more human-like way to check the task

[flibbertigibbet-Y]

I can see that tasks are checked by either reading the system setting status or reading database information.

Just out of curiosity have you ever considered to check the if the tasks have been properly finished by comparing the UI elements status ?
I guess this is a more human like way

[crawles]

Yes, we do check UI elements for some tasks. For example, the Clock app we use UY elements because there’s no readily available API https://github.com/google-research/android_world/blob/47f8cfb0be07ff5f929b2fbfe3dc52107846c3bd/android_world/task_evals/single/clock.py#L152 We prefer to use system state from internal sources, when available because we found them to be more durable than checking UI elements Chris

…

On Wed, Feb 19, 2025 at 9:45 PM Flibbertigibbet ***@***.***> wrote: I can see that tasks are checked by either reading the system setting status or reading database information. Just out of curiosity have you ever considered to check the if the tasks have been properly finished by comparing the UI elements status ? I guess this is a more human like way — Reply to this email directly, view it on GitHub <#223>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABG26SOPEC5YF27A7SX4AMD2QU6TBAVCNFSM6AAAAABXPVBVRSVHI2DSMVQWIX3LMV43ASLTON2WKOZSHA3DIOJSGM3TINQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> [image: flibbertigibbet-Y]*flibbertigibbet-Y* created an issue (google-research/android_world#223) <#223> I can see that tasks are checked by either reading the system setting status or reading database information. Just out of curiosity have you ever considered to check the if the tasks have been properly finished by comparing the UI elements status ? I guess this is a more human like way — Reply to this email directly, view it on GitHub <#223>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABG26SOPEC5YF27A7SX4AMD2QU6TBAVCNFSM6AAAAABXPVBVRSVHI2DSMVQWIX3LMV43ASLTON2WKOZSHA3DIOJSGM3TINQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>