Fact-checking, NLI, generalization, interpretability, BERTology, holistic evaluation of neural models