Yes, you need to exponentiate the RMSE to interpret it correctly in the original scale of the label variable.

No, you do not need to exponentiate the RMSE because it is already in the original scale of the label variable.

Yes, you need to exponentiate the RMSE to interpret it correctly, but only if you are using a specific loss function.

No, you do not need to exponentiate the RMSE because it is not affected by the log transformation of the label variable.