This is the *best performing* model used in the paper: "End-to-end Training For Financial Report Summarization" 


https://www.aclweb.org/anthology/2020.fnp-1.20/