Tu Vu sur Twitter : "Enormous LMs like GPT-3 exhibit impressive few-shot performance, but w/ self-training a BERT base sized model can achieve much better results!
Tags:
About This Document
File info
Documents with similar tags (experimental)