Since it's a transformer model, I use it to... "transform" the data. - Change raw citation into bibtex entry - Fix spelling mistakes - convert csv file into json (though you can do it without employing trillion parameter model) etc.