Can NLP Models Correctly Reason Over Contexts that Break the Common Assumptions?

Neeraj Varshney, Mihir Parmar, Nisarg Patel +4

Pre-training on large corpora of text enables the language models to acquire a vast amount of factual and commonsense knowledge which allows them to achieve remarkable performance…