BioinformaticsbioRxivPreprint
Robustness to nuisance perturbations enables unsupervised evaluation of single-cell foundation models
This paper evaluates single-cell foundation models using an unsupervised framework that assesses their robustness to nuisance perturbations, revealing significant differences in structural stability between models ranked similarly by standard benchmarks.
DOI 10.64898/2026.08.22.746357