A paper I reviewed for #emnlp2020 got two low scores because: "I worry that this paper is the outcome of a lab with insufficient resources to run meaningful, representative experiments on BERT training." I strongly argued against this sentiment & I'm happy to see it accepted :)
