=========================================
 Split-Leakage Gate (model-validation)
=========================================
  rows=14  subjects=12  partitions={'test': 4, 'train': 6, 'val': 4}  seed=42
| Check | Severity | Detail |
|---|---|---|
| PATIENT_OVERLAP | Major | 2 of 12 subjects appear in >= 2 partitions (e.g. 'P03' in test/train); the same patient in train and test inflates every metric. Offenders: P03, P07 |

MAJOR candidate: 1 split-integrity issue(s).
