GPT-3 Models are Poor Few-Shot Learners in the Biomedical Domain

1 code implementation • 6 Sep 2021 • Milad Moradi, Kathrin Blagec, Florian Haberl, Matthias Samwald

However, in-domain pretraining seems not to be sufficient; novel pretraining and few-shot learning strategies are required in the biomedical NLP domain.

Few-Shot Learning Language Modelling +1