mrqa/MRQA-Shared-Task-2019

Is that possible to provide baseline performance on in-domain dev set?

boyang-nlp opened this issue · 3 comments

Is that possible to provide the performance of the pytorch baseline (with BERT_base and BERT_large) on in-domain dev set? Then, each participant does not need to re-run it if they choose to develop their own system with other framework. Thanks.

Hi! Yes, we will release both the trained baseline models as well as the results on the in-domain and out-of-domain dev sets next week (May 13). Check back soon!

Thanks!

Baseline models and results are now available. See the baseline directory.