Deep learning perturbation models can outperform baselines on calibrated metrics

T1Covered by 1 source · first reported by Nature / Science / Nature Biotechnology at 01 Oct, 00:00 UTC

Nature Biotechnology, Published online: 01 October 2026; doi:10.1038/s41587-026-03307-w Calibration-aware evaluation of perturbation models shows that deep learning models can outperform baselines.

Sources

• T1Deep learning perturbation models can outperform baselines on calibrated metricsNature / Science / Nature Biotechnology01 Oct, 00:00