I was wondering can you release how you produced the evaluation set (data/*.txt)? I.e. the setup of the controlled text generation model that was used to produce the text. I'm interested in using more generations of the same setup, without additional human annotated scores. Thanks a lot!
I was wondering can you release how you produced the evaluation set (
data/*.txt)? I.e. the setup of the controlled text generation model that was used to produce the text. I'm interested in using more generations of the same setup, without additional human annotated scores. Thanks a lot!