Six Open Questions in Machine-Learned Interatomic Potential Foundation Models
7 affiliations
- 1Department of Chemistry , Imperial College London , South Kensington Campus , London , SW7 2AZ , UK
- 2The Thomas Young Centre , London WC1E 6N , UK
- 3Department of Chemistry , University College London , Kathleen Lonsdale Building , Gower Pl , London , WC1E 6BS , UK
- 4Department of Physics , Imperial College London , South Kensington Campus , London , SW7 2AZ , UK
- 5Department of Chemical Engineering , University College London , Roberts Building , Torrington Pl , London , WC1E 7JE , UK
- 6College of Chemistry , University of California Berkeley , CA 94720-146 , USA
- 7Laboratory of Computational Science and Modeling , Institute of Materials , École Polytechnique Fédérale de Lausanne , 1015 Lausanne , Switzerland
Abstract
Machine-learned interatomic potentials (MLIPs) have had a profound impact on molecular modelling in recent years, promising to resolve the long-standing tension between the scale and accuracy of simulations. There has been a proliferation of new models and designs, and recently the paradigm of ``foundational'' MLIPs has become prevalent. Broadly speaking, foundation models are trained on large diverse datasets and promise to work well for new systems with minimal updates required. However, in such a new and fast moving field, there are many unanswered questions. In this article, we set out to articulate and explore what we see as the most important among these questions. We start by developing a working definition for foundational MLIPs and use this definition to frame the subsequent open questions. Despite the rapid progress in the field of MLIP models, we believe that these are fundamental questions which will continue to define cutting edge research in MLIPs in the years to come.
Related
- Accelerating Mechanistic Understanding and Prediction of Chemical Selectivity with Foundation Machine-Learned Interatomic Potentials
- A Machine-Learned Interatomic Potential Free-Energy Surface for the First Step of MIL-101(Cr) Secondary Building Unit Formation
- GRACE-OFF: A Machine-Learned Interatomic Potential for Organic Liquids Using the GRACE Architecture.
- Accurate and Rapid Vibrational Spectra of Molecular Crystals from Machine-Learned Interatomic Potentials
- Neural Constrained Charge Equilibration for Extracting Charge Transport Parameters from Machine Learned Interatomic Potentials
BibTeX
@misc{creed2026open,
title = {{Six Open Questions in Machine-Learned Interatomic Potential Foundation Models}},
author = {Isabel Creed and Tim Rein and Ingvars Vitenburgs and Wojciech G. Stark and Viktor Ellingsson and Assoujaa Ismail and Guangyu Liu and Yuchen Lou and Bradley A. A. Martin and Cyprien Bone and Matthew A. H. Walker and Mueen Taj and Shirui Wang and Kelvin Wong and Ruiqi Wu and Prakriti Kayastha and Bingqing Cheng and Aditi S. Krishnapriyan and Michele Ceriotti and Marcel F. Langer and Jarvist Moore Frost and Alex M. Ganose and Venkat Kapil and Keith T. Butler},
howpublished = {arXiv (Cornell University)},
year = {2026},
month = jun,
doi = {10.48550/arxiv.2606.07327},
url = {https://arxiv.org/abs/2606.07327},
note = {CC BY},
}
RIS
TY - UNPB
TI - Six Open Questions in Machine-Learned Interatomic Potential Foundation Models
AU - Isabel Creed
AU - Tim Rein
AU - Ingvars Vitenburgs
AU - Wojciech G. Stark
AU - Viktor Ellingsson
AU - Assoujaa Ismail
AU - Guangyu Liu
AU - Yuchen Lou
AU - Bradley A. A. Martin
AU - Cyprien Bone
AU - Matthew A. H. Walker
AU - Mueen Taj
AU - Shirui Wang
AU - Kelvin Wong
AU - Ruiqi Wu
AU - Prakriti Kayastha
AU - Bingqing Cheng
AU - Aditi S. Krishnapriyan
AU - Michele Ceriotti
AU - Marcel F. Langer
AU - Jarvist Moore Frost
AU - Alex M. Ganose
AU - Venkat Kapil
AU - Keith T. Butler
PY - 2026
DA - 2026/06/05
PB - arXiv (Cornell University)
DO - 10.48550/arxiv.2606.07327
UR - https://arxiv.org/abs/2606.07327
AB - Machine-learned interatomic potentials (MLIPs) have had a profound impact on molecular modelling in recent years, promising to resolve the long-standing tension between the scale and accuracy of simulations. There has been a proliferation of new models and designs, and recently the paradigm of ``foundational'' MLIPs has become prevalent. Broadly speaking, foundation models are trained on large diverse datasets and promise to work well for new systems with minimal updates required. However, in such a new and fast moving field, there are many unanswered questions. In this article, we set out to articulate and explore what we see as the most important among these questions. We start by developing a working definition for foundational MLIPs and use this definition to frame the subsequent open questions. Despite the rapid progress in the field of MLIP models, we believe that these are fundamental questions which will continue to define cutting edge research in MLIPs in the years to come.
SN - 2331-8422
ER -