Measuring feature importance for additional regressors

Hi,

In one the last releases you've added feature with external regressors. Is there any possibility to estimate their impact on the target variable **except** training models with them and without them and comparing prediction accuracy on holdout set?

Thanks.