JieyuZ2/wrench

Wrong Data Statistics of the SemEval Dataset?

davidpig opened this issue · 1 comments

davidpig commented

Dear authors,

thanks a lot for the nice repo!

I noticed that both the WRENCH paper and this repo report the following data statistics on the SemEval dataset:
number of samples in train/validation/test： 1749/200/692

However, I found the data in the link you provided (Google Drive) have 1749, 178, and 600 samples for training, validation, and test. Maybe I misunderstood something?

Looking forward to hearing from you.

JieyuZ2 commented

Hey,

Thanks for pointing this out- it should be a typo and just fixed the table in readme

  • 👍1
  • 😄1