Nature paper: deep learning perturbation models can beat baselines on calibration

On October 1, 2026, Nature's Machine Learning section published a study titled “Deep learning perturbation models can outperform baselines on calibrated metrics.” The study concludes that deep learning perturbation models can outperform baseline methods on calibrated metrics. The report also states that no abstract was provided by the source, so the specific models, datasets, and experimental setup described in the paper cannot be verified. The matter currently rests at the publication of the paper, with no further developments available.