Unable to Reproduce Results on Arxiv Dataset

[wangminllw]

I am experiencing difficulties reproducing the results on the Arxiv dataset using the GNNSAFE setting. While I have successfully reproduced results on other datasets, the Arxiv dataset results I obtained significantly differ from those reported in your paper. Below are the results I achieved: Loss: 1.0974，AUROC: 67.68%，AUPR: 76.26%，FPR95: 88.00%，Test Score: 52.59%

These results are quite different from those reported in your article. Could you provide guidance or insights into what might be causing this discrepancy and how I might resolve it?

Thank you for your assistance.

[qitianwu]

According to our experience, the possible reasons could be you using different package versions or computation machines. Particularly, the package version could sometimes lead to large variation. Have you followed the requirements.txt for installing the environment?