princeton-nlp / LM-BFF

[ACL 2021] LM-BFF: Better Few-shot Fine-tuning of Language Models https://arxiv.org/abs/2012.15723

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

A little question about the evaluation

shadowkiller33 opened this issue · comments

Hi! Thanks for your wonderful work!
I got a little question about the zero-shot evaluation on text classification. When doing argmax to get the predicted label, are we going to argmax on the whole vocab or on the specific label set (e.g., pos/neg in sentiment analysis)

Thanks a LOT

Hi,

We did it only on the specific label set for both zero-shot and few-shot settings.