When rows repeat a person, the split changes the question
Six practice rows from participants A, B, and C (two sessions each). Compare a row split and a person split. Shapes ▣/▢ mark train vs test as well as color.
First-attempt score: 0 / 5 (0 answered)
Row split
Session 1 → train, session 2 → test.
| Row | Set |
|---|
- Train IDs:
- Test IDs:
- IDs in both sets:
Person split
A and B entirely in train; C entirely in test.
| Row | Set |
|---|
- Train IDs:
- Test IDs:
- IDs in both sets:
Five questions
Your first choice on each question counts toward the score. After you submit, that answer locks.
Miss journal
Reflection (not scored)
Reveal the explanation when you are ready, then self-rate — it does not affect your score.
Person-level splits keep every row from one participant in the same set, which is the usual choice when you need to score on a new person. Row splits that scatter sessions can inflate scores because the model sees the same identity in both train and test.