The caret library in R has a hyper-parameter 'selectionFunction' inside trainControl(). It's used to prevent over-fitting models using Breiman's one standard error rule, or tolerance, etc.
Does mlr have an equivalent? If so, which function is it within?