The evaluation results as well as leaderboards look very different

[zhimin-z]

The original Arxiv paper:

The GitHub Readme:

[auhowielau]

The paper shows the results on the test sets. Since the test set data hasn't been released, we provide the results on the validation sets for reference in Readme.

[zhimin-z]

The paper shows the results on the test sets. Since the test set data hasn't been released, we provide the results on the validation sets for reference in Readme.

This should be clarified in the documentation, thanks!