This is the SciBERT pretrained language model further fine-tuned on masked language modeling and cite-worthiness detection on the CiteWorth dataset. Note that this model should be used for further fine-tuning on downstream scientific document understanding tasks.