sev777 opened this issue 7 months ago · 0 comments
Why are only binary classification tasks tested for NLP? Shouldn't generation tasks be more suitable as generative models?