Specifically, it was trained with two simultaneous self-supervised objectives, language modeling and link prediction, that encourage deep bidirectional reasoning over text and knowledge graphs. DRAGON ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results