Open access
Sep 2026
Cell-level random splits leak group-owned answers in single-cell benchmarks
This work shows how to estimate leakage before training any model, from two properties of the data: exposure, the fraction of test cells with relatives in training, and retrievability, how often a nearest-neighbor search returns such a relative rather than an unrelated cell.
Cang Hu, Sha Sun
· bioRxiv · 0 citations