Ranked 15 / 326 in the Tianchi NLP Learning Competition: Model Generalization Challenge for Pretrained Models.