1 paper
Sriram Balasubramanian, Naman Jain, Gaurav Jindal +2
We evaluate named entity representations of BERT-based NLP models by investigating their robustness to replacements from the same typed class in the input. We highlight that on sev…