[ ] If the data has natural groups (patients, users, companies), all observations from the same group are in the same split - [ ] No duplicate or near-duplicate rows exist across train and test