UniFFBench: Evaluating universal machine learning force fields for experimental measurements

Machine Learning


  • Duval, A. et al. A hitchhiker’s guide to geometric GNNs for 3D atomic systems. Preprint available at https://arxiv.org/abs/2312.07511 (2023).

  • Bihani, V. et al. Egraffbench: Evaluating equivariant graph neural network force fields for atomistic simulations. digit. Discob. 3759–768 (2024).

    Article Google Scholar

  • Bhattatia, I. et al. Basic model of atomic materials chemistry. J. Chem. Physics. 163184110 (2025).

    Article Google Scholar

  • Musaelian, A. et al. Learning locally equivariant representations for large-scale nuclear science. nut. common. 14579 (2023).

    Article Google Scholar

  • Fu, X. et al. Force alone is not enough: Benchmarking and critical evaluation of machine learning force fields with molecular simulations. Preprint available at https://arxiv.org/abs/2210.07237 (2023).

  • Miret, S., Lee, KLK, Gonzales, C., Nassar, M. & Spellings, M. Open MatSci ML Toolkit: A flexible framework for machine learning in materials science. Preprint available at https://arxiv.org/abs/2210.17484 (2023).

  • Saal, JE, Kirklin, S., Aykol, M., Meredig, B. & Wolverton, C. High-throughput density functional theory materials design and discovery: Open Quantum Materials Database (OQMD). JOM 651501–1509 (2013).

    Article Google Scholar

  • Hautier, G., Jain, A. & Ong, SP From the computer to the laboratory: Materials discovery and design using ab initio calculations. J. Mater. Science. 477317–7340 (2012).

    Article Google Scholar

  • Merchant, A. et al. Scaling deep learning for materials discovery. nature 62480–85 (2023).

    Article Google Scholar

  • Axelrod, S. and more: Materials design using machine learning and atomic simulation. Compliant meter. resolution 3343–357 (2022).

    Article Google Scholar

  • Formerly, EC-Y. Others. Fundamental model for atomistic simulation of chemistry and materials. nut. Rev. Chem. 10212–230 (2026).

    Article Google Scholar

  • Schmidt, J. et al. Improving machine learning models in materials science through large datasets. meter. Today is physics. 48101560 (2024).

    Article Google Scholar

  • Lee, KLK et al. MatSciML: An extensive multitasking benchmark for solid-state materials modeling. Preprint available at https://arxiv.org/abs/2309.05934 (2023).

  • Deringer, VL, Caro, MA & Csányi, G. Interatomic potentials for machine learning as a new tool in materials science. Advanced meter. 311902765 (2019).

    Article Google Scholar

  • Friederich, P., Häse, F., Proppe, J., Aspuru-Guzik, A. The potential of machine learning for next generation materials simulation. nut. meter. 20750–761 (2021).

    Article Google Scholar

  • Zeni, C. et al. Generative models for inorganic materials design. nature 639624–632 (2025).

    Article Google Scholar

  • Deng, B. et al. CHGNet is a pre-trained universal neural network with the potential for atomistic modeling based on charge information. nut. Mach. intelligence. 51031–1041 (2023).

    Article Google Scholar

  • Chen, C. & Ong, SP Universal graph deep learning interatomic potentials for the periodic table. nut. Calculate. Science. 2718–728 (2022).

    Article Google Scholar

  • Yang, H et al. MatterSim: Deep learning atomistic models across elements, temperature, and pressure. Preprint available at https://arxiv.org/abs/2405.04967 (2024).

  • Park, Y., Kim, J., Hwang, S. & Han, S. A scalable parallel algorithm for interatomic potential of graph neural networks in molecular dynamics simulations. J. Chem. Theoretical calculation. 204857–4868 (2024).

    Article Google Scholar

  • Neumann, M. et al. ORB: The potential of fast, scalable neural networks. Preprint available at https://arxiv.org/abs/2410.22570 (2024).

  • Barroso-Luque, L. et al. Open Materials 2024 (OMat24) Inorganic materials dataset and models. Preprint available at https://arxiv.org/abs/2410.12771 (2024).

  • Levine, DS et al. Open Molecules 2025 (OMol25) dataset, evaluation, and models. Preprint available at https://arxiv.org/abs/2505.08762 (2025).

  • Chanusot, L. et al. Open Catalyst 2020 (OC20) dataset and community challenges. ACS Catalog. 116059–6072 (2021).

    Article Google Scholar

  • Liebesel, J. et al. A framework for evaluating crystal stability prediction by machine learning. nut. Mach. intelligence. 7836–847 (2025).

    Article Google Scholar

  • Miret, S., Lee, KLK, Gonzales, C., Mannan, S. & Krishnan, NM DFT orbital energy and force regression is not sufficient for universal machine learning interatomic potentials. Preprint available at https://arxiv.org/abs/2502.03660 (2025).

  • Wood, B.M. et al. UMA: Family of universal models of atoms. Advanced neural information processes. system. 38129391–129427 (2026).

    Google Scholar

  • Chiang, Y. et al. MLIP Arena: Improving the fairness and transparency of machine learning interatomic potentials through an open and accessible benchmarking platform. Advanced neural information processes. system. 38 (2026).

  • Fu, X. et al. Force alone is not enough: Benchmarking and critical evaluation of machine learning force fields with molecular simulations. (2023).

  • Fu, X. et al. Learn smooth and expressive interatomic potentials for predicting physical properties. Preprint available at https://arxiv.org/abs/2502.12147 (2025).

  • Zaki, M., Mausam, J. & Krishnan, NMA MaScQA: Exploring materials science knowledge about large-scale language models. digit. Discob. 3313–327 (2024).

    Article Google Scholar

  • Miret, S. & Krishnan, NMA Enabling large-scale language models for real-world materials discovery. nut. Mach. intelligence. 7991–998 (2025).

  • Arampara, N. et al. Exploring the limitations of multimodal language models in chemistry and materials research. nut. Calculate. Science. 5952–961 (2025).

    Article Google Scholar

  • Gonzales, C., Fuemmeler, E., Tadmor, EB, Martiniani, S. & Miret, S. A universal machine learning interatomic potential benchmark for structural relaxation. open review https://openreview.net/pdf?id=fNyXCCZ0g6 (2024).

  • Tran, R. et al. Open Catalyst 2022 (OC22) dataset and challenges for oxide electrocatalysts. ACS Catalog. 133066–3084 (2023).

    Article Google Scholar

  • Warford, T., Thiemann, FL & Csányi, G. Better without U: The impact of selective Hubbard U correction on basic MLIP. Mach. learn. 7035033 (2026).

    Google Scholar

  • Gangan, AS, Cubuk, ED, Schoenholz, SS, Bauchy, M. & Krishnan, NMA End-to-end differentiable atomistic simulation force field optimization. J. Chem. Theoretical calculation. twenty one5867–5879 (2025).

    Article Google Scholar

  • Thaler, S. & Zavadlav, J. Learning neural network likelihoods from experimental data by weighting differentiable trajectories. nut. common. 126884 (2021).

    Article Google Scholar

  • Downs, RT & Hall-Wallace, M. American Mineralogist’s Crystal Structure Database. morning. mineral. 88247–250 (2003).

    Google Scholar

  • T. Hahn, U. Shmueli, JCW Arthur international crystallography table Vol. 1 (Leidel Dordrecht, 1983).

  • Vaclav, H. & Zimmermann, H. Hermann Morgan Symbol (Wiley Online Library, 2016).

  • Arroyo, Michigan International Crystallography Table: Crystallographic Symmetries (John Wiley & Sons, 2021).

  • Larsen, AH et al. Atomic Simulation Environment – A Python library for manipulating atoms. J. Phys. Condense. Case 29273002 (2017).

    Article Google Scholar

  • Material Properties Open Database (MPOD). MPOD http://mpod.cimav.edu.mx/ (accessed August 1, 2025).

  • Bitzek, E., Koskinen, P., Gähler, F., Moseler, M., Gumbsch, P. Structural relaxation made simple. Physics. Pastor Rhett. 97170201 (2006).

    Article Google Scholar

  • Berendsen, HJC, van Postma, JPM, Van Gunsteren, WF, DiNola, ARHJ & Haak, JR Molecular dynamics with coupling to an external bus. J. Chem. Physics. 813684–3690 (1984).

    Article Google Scholar

  • Matsumoto, Y. Overview of Morse Theory Vol. 208 (Mathematical Association of America, 2002).

  • Mannan, S. et al. Evaluation of universal machine learning force fields for experimental measurements [Dataset]. Zenodo https://doi.org/10.5281/zenodo.16733258 (2026).

  • Mannan, S., Bihani, V., Gonzales, C., Maheshwar, A. M3RG-IITD/UniFFBench: UniFFBench-v1.0.1 (v1.0.1). Zenodo https://doi.org/10.5281/zenodo.20165458 (2026).



  • Source link