Tweeted By @yoavgo
claims about "human performance" in this context are very misleading, they are not measuring the human performance on QA but on the SQUAD task. They also compare against a very specific set of humans. https://t.co/RBKvXzpNBG
— (((ل()(ل() 'yoav)))) (@yoavgo) March 28, 2019