On a Benefit of Masked Language Model Pretraining: Robustness to Simplicity Bias.

Ting-Rui Chiang

Browse the full IJCNLP paper archive.