Comment by startupsfail

The results are repeatable. Models are performing with predictable error rates on the tasks that these models had been trained and tested.