In July 2019, Prof. Zhao Hai’s team ranked 1st for single model on SQuAD2.0 leaderboard, a flagship leaderboard in the field of machine reading comprehension with their model outperforming human comprehension. Recently, the team won the 1st place for both single and ensemble models on the leaderboard agains strong competitors.
Preprint of the paper can be seen at:https://arxiv.org/abs/2001.09694