About performance of w/o Adatper

[Xianqi-Zhang]

Hi,
Thank you very much for your insightful sharing.

In TABLE V, the performance of the model w/o the adapter seems to contribute only marginally. It also appears that comparisons with other residual RL methods—both residual dynamics and residual actions—are not included.

Could you kindly provide some guidance on the advantages of AnyAdapter over these methods? Additionally, have you attempted such comparisons, or do you plan to conduct related experiments in the future?

Best Regards.