We propose a new pre-trained language model called MacBERT that mitigates the gap be- tween the pre-training and fine-tuning stage by masking ...
確定! 回上一頁