The per-study folders read experiment outputs (artifacts/, runs/, scores/) that this repository does not ship. They are kept so that each reported analysis has its code; they run once those outputs ...