About the results on the CB dataset

THUDM / P-tuning-v2

An optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks

Apache License 2.0

1.96k stars 198 forks source link

About the results on the CB dataset #35

Closed Dicer-Zz closed 2 years ago

Dicer-Zz commented 2 years ago

Why does the effect of fine-tuning on the CB dataset using Bert-large in this paper exceed the SuperGLUE baseline by more than ten absolute points (83.6 -> 94.6)? And the results on other datasets are not much different.

Xiao9905 commented 2 years ago

@Dicer-Zz Hi,

We report the results on the development set of SuperGLUE (Page 16, Table 4).

Dicer-Zz commented 2 years ago

Thank you for your quick reply! I get it.