search

google-research / FLAN

Apache License 2.0
1.47k stars 154 forks source link

Thank you for your efforts. This is a great job！I have a problem. Why is there only 70 tasks in this FLAN2021_submix_original, and it seems that there are over 140 in the paper? #82

Open flyinghpluo opened 1 year ago

flyinghpluo commented 1 year ago

Thank you for your efforts. This is a great job！I have a problem. Why is there only 70 tasks in this FLAN2021_submix_original, and it seems that there are over 140 in the paper?

shayne-longpre commented 1 year ago

@flyinghpluo Thank you for your question.

The full list of released datasets/tasks is here. A couple were omitted, as described in the second "NB #2" here.

I think the discrepancy you're referring to is related to how we define and count "dataset" or "task" -- the convention is different across papers. Included in this repository are input inverted versions of most of the datasets (e.g. [question --> answer] becomes [answer --> question]), which we count as an additional task. But, with the exception of a couple datasets, all tasks are in this repository.