I have a question about the comparison of GROVER and your model.
In your paper (https://www.nature.com/articles/s42004-025-01585-0), for example you compare the RMSE of GROVER and your model, and RMSE of GROVER is 2.272 in FreeSolv dataset in your manuscript.
But original paper of GROVER (https://proceedings.neurips.cc/paper_files/paper/2020/file/94aef38441efa3380a3bed3faf1f9d5d-Paper.pdf), RMSE is 1.544.
Both papers seem to use the same ``scaffold'' splitting of dataset in evaluation, but why these RMSEs are different? Are the original paper of GROVER and your paper different setting of scaffold datasets? Where does RMSE 2.272 of GROVER come from or your own experiment?
I have a question about the comparison of GROVER and your model.
In your paper (https://www.nature.com/articles/s42004-025-01585-0), for example you compare the RMSE of GROVER and your model, and RMSE of GROVER is 2.272 in FreeSolv dataset in your manuscript.
But original paper of GROVER (https://proceedings.neurips.cc/paper_files/paper/2020/file/94aef38441efa3380a3bed3faf1f9d5d-Paper.pdf), RMSE is 1.544.
Both papers seem to use the same ``scaffold'' splitting of dataset in evaluation, but why these RMSEs are different? Are the original paper of GROVER and your paper different setting of scaffold datasets? Where does RMSE 2.272 of GROVER come from or your own experiment?