This came up today and was confusing for a few people:

library(mlr3)
library(mlr3viz)
lrn_list = list(
  lrn("classif.log_reg", predict_type = "prob"),
  lrn("classif.kknn", predict_type = "prob", k = 3),
  lrn("classif.kknn", predict_type = "prob", k = 10)
)

bm_design = benchmark_grid(task = tsk("sonar"), resamplings = rsmp("cv", folds = 10), learners = lrn_list)
bench_result = benchmark(bm_design)
autoplot(bench_result)

Both knn learners are grouped together. I know that this can be fixed by setting ids but this is still easy to overlook as there are now 2 different learners combined in a single boxplot.

Imo. There should be 3 boxplots with two of them having the same name. (maybe do the faceting by the learner hash and overwrite the labels with the learner id afterward?)