English

Reducing cross-sample prediction churn in scientific machine learning

Machine Learning 2026-05-14 v1 Materials Science Chemical Physics

Abstract

Scientific machine learning reports predictive performance. It does not report whether the same prediction would survive a different draw of training data. Across 99 chemistry benchmarks, two classifiers trained on independent bootstraps of the same training set agree on aggregate accuracy to within 1.34.21.3\text{--}4.2 percentage points but disagree on the class label of 8.021.8%8.0\text{--}21.8\% of test molecules. We call this gap \emph{cross-sample prediction churn}. The standard parameter-side techniques (deep ensembles, MC dropout, stochastic weight averaging) do not reduce this gap; two data-side methods do. The first is KK-bootstrap bagging, which cuts the rate 4054%40\text{--}54\% on every dataset at no accuracy cost (K×K{\times}-ERM compute). The second is \emph{twin-bootstrap}, our proposal: two networks trained jointly on independent bootstraps with a sym-KL consistency loss between their predictions, which at matched 2×2{\times}-ERM compute reduces churn a further median 45%45\% beyond bagging-K=2K{=}2. Cross-sample prediction churn deserves a column alongside predictive performance in scientific-ML benchmark reports, because without it the parameter-side and data-side methods are indistinguishable on the metric they actually differ on.

Keywords

Cite

@article{arxiv.2605.13826,
  title  = {Reducing cross-sample prediction churn in scientific machine learning},
  author = {Gordan Prastalo and Kevin Maik Jablonka},
  journal= {arXiv preprint arXiv:2605.13826},
  year   = {2026}
}