This paper uses gold entity information rather than the predicted entities to evaluate the model.
It is inappropriate to say that you achieved the best results on NYT and WebNLG, since previous works (such as HBT) did not use the gold entity information when testing.
This paper uses gold entity information rather than the predicted entities to evaluate the model. It is inappropriate to say that you achieved the best results on NYT and WebNLG, since previous works (such as HBT) did not use the gold entity information when testing.