Whenever I read papers, I always wonder what evaluations are good. I know that it depends on machine learning tasks, and a result can be changed by splitting data. In the or
