Closed RicardoUsbeck closed 2 years ago
In some ways, it is meaningful as we have already tracked the evaluation results of MetaQA and PathQuestions datasets in the leaderboard, both of which are used in this paper.
The evaluation results of these two datasets have already been updated.
close this issue if merged
I do not think so, since they only give evaluation on subsets (1-hop, 2-hop, 3-hop) of kgqa datasets. And we do not want that, right?