62% accuracy, uses transformer models.

I started reviewing the paper with an eye toward why the accuracy was only 62%, but did not find their training information in the time provided.