← Back to all news
arXiv Machine Learning By Amir Masoud Nourollah, Irtaza Khalid, Stefano Leoni, Steven Schockaert

Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials

Read the original on arXiv Machine Learning →

The paper introduces a benchmark to evaluate compositional generalisation in machine learning interatomic potentials. It comprises four tasks that test models on unseen molecules, designed so that learning underlying chemical principles should enable generalisation. Experiments show that current state‑of‑the‑art models struggle, with out‑of‑distribution errors often an order of magnitude higher than in‑distribution errors, even for foundation models pre‑trained on millions of molecules.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv Machine Learning
Jul 1

ElemeNet: Multiscale Molecular Machine Learning with Uncertainty Quantification Across the Periodic Table

arXiv:2606. 30961v1 Announce Type: cross Abstract: Advances in deep learning architectures and representations have enabled ML-driven chemical property prediction, but state-of-the-art (SOTA) models have remained largely confined to independent codebases and lack support for diverse chemical species.

By Jacob W. Toney, Samir Darouich, Yiran Wang, Aaron G. Garrison, Johannes K\"astner, Heather J. Kulik
llmsbenchmarks
More like this →
arXiv AI
Sep 17

BOOM: Benchmarking Out-Of-distribution Molecular Property Predictions of Machine Learning Models

BOOM is a new benchmark for evaluating out‑of‑distribution (OOD) molecular property predictions in machine learning. It provides chemically‑informed tests across common property prediction tasks and assesses over 150 model‑task combinations. The study shows that current models, including chemical foundation models, struggle to generalize OOD, with the best model still exhibiting three times higher error than in‑distribution predictions.

By Evan R. Antoniuk, Shehtab Zaman, Tal Ben-Nun, Peggy Li, James Diffenderfer, Busra Sahin, Obadiah Smolenski, Everett Grethel, Tim Hsu, Anna M. Hiszpanski, Kenneth Chiu, Bhavya Kailkhura, Brian Van Essen
benchmarkssafety
More like this →
arXiv AI
Oct 15

Can Machine Learning Potentials Be More Accurate Than the Methods They Are Trained On?

The article discusses how machine learning potentials (MLPs) can surpass the accuracy of the quantum‑chemical methods they are trained on by generalizing across a third space—different quantum‑chemical methods. It proposes the development of method‑transferable MLPs that predict properties at unseen methods, demonstrating that such models can outperform any of their training methods in both density‑functional and wavefunction‑based theories. The authors argue that pursuing accuracy beyond the training methods could enable universal MLPs to exceed single‑method models, influencing dataset construction and model evaluation to push chemical simulation limits.

By Pavlo O. Dral, Yi-Fan Hou, Yuxinxin Chen, Hassan Nawaz, Mateusz Witkowski, Szymon \'Smiga, Ireneusz Grabowski
More like this →