This is a BERT-large model (bert-large-uncased-whole-word-masking) fine-tuned on the CUAD dataset from CUAD: An Expert-Annotated NLP Dataset for Legal Contract Review (Hendrycks et al., 2021), with the BertforQuestionAnswering model architecture.

The questions ask for information often found in contracts; the model would return the relevant text string and its starting index in the given document if the information exists. The CUAD dataset is in SQuAD 2.0 format.

For details of the dataset and usage of the relevant training/testing scripts, check out the paper and their Github repo.