mlsanityData to 7 Oct 2026

Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction

arXiv 2502.12147 ↗First posted 17 Feb 2025physics.comp-ph cs.LGcited by 79

Not on the map. Why?

Abstract (arXiv)

Machine learning interatomic potentials (MLIPs) have become increasingly effective at approximating quantum mechanical calculations at a fraction of the computational cost. However, lower errors on held out test sets do not always translate to improved results on downstream physical property prediction tasks. In this paper, we propose testing MLIPs on their practical ability to conserve energy during molecular dynamic simulations. If passed, improved correlations are found between test errors and their performance on physical property prediction tasks. We identify choices which may lead to models failing this test, and use these observations to improve upon highly-expressive models. The resulting model, eSEN, provides state-of-the-art results on a range of physical property prediction tasks, including materials stability prediction, thermal conductivity prediction, and phonon calculations.

Key facts. Cited by 79 papers in the map (data through 7 Oct 2026). Builds on 40 of its 93 references that are in the map. First posted on arXiv 17 Feb 2025.

Builds on

40 of 93 references.

References · 40

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2016Layer Normalizationcited by 5,579
  3. 2017Neural Message Passing for Quantum Chemistrycited by 3,460
  4. 2016Group Equivariant Convolutional Networkscited by 1,154
  5. 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 735
  6. 2021E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentialscited by 701
  7. 2017SchNet: A continuous-filter convolutional neural network for modeling quantum interactionscited by 569
  8. 2020Directional Message Passing for Molecular Graphscited by 556
  9. 2022MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fieldscited by 467
  10. 2021Equivariant message passing for the prediction of tensorial properties and molecular spectracited by 438
  11. 20183D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Datacited by 348
  12. 2021GemNet: Universal Directional Graph Neural Networks for Moleculescited by 342
28 more
  1. 2022Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphscited by 258
  2. 2022Learning Local Equivariant Representations for Large-Scale Atomistic Dynamicscited by 254
  3. 2020The Open Catalyst 2020 (OC20) Dataset and Community Challengescited by 246
  4. 2020Machine Learning Force Fieldscited by 204
  5. 2023EquiformerV2: Improved Equivariant Transformer for Scaling to Higher-Degree Representationscited by 201
  6. 2022e3nn: Euclidean Neural Networkscited by 171
  7. 2021Spherical Message Passing for 3D Graph Networkscited by 166
  8. 2023Reducing SO(3) Convolutions to SO(2) for Efficient Equivariant GNNscited by 113
  9. 2021SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effectscited by 103
  10. 2022The Design Space of E(3)-Equivariant Atom-Centered Interatomic Potentialscited by 100
  11. 2022The Open Catalyst 2022 (OC22) Dataset and Challenges for Oxide Electrocatalystscited by 95
  12. 2022Forces are not Enough: Benchmark and Critical Evaluation for Machine Learning Force Fields with Molecular Simulationscited by 92
  13. 2022SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentialscited by 80
  14. 2024Orb: A Fast, Scalable Neural Network Potentialcited by 63
  15. 2021ForceNet: A Graph Neural Network for Large-Scale Quantum Calculationscited by 58
  16. 2024Does equivariance matter at scale?cited by 49
  17. 2022Spherical Channels for Modeling Atomic Interactionscited by 47
  18. 2024The Importance of Being Scalable: Improving the Speed and Accuracy of Neural Network Interatomic Potentials Across Chemical Domainscited by 40
  19. 2024Cartesian atomic cluster expansion for machine learning interatomic potentialscited by 31
  20. 2020Differentiable Molecular Simulations for Control and Learningcited by 25
  21. 2022AdsorbML: A Leap in Efficiency for Adsorption Energy Calculations using Generalizable Machine Learning Potentialscited by 25
  22. 2024The dark side of the forces: assessing non-conservative force models for atomistic machine learningcited by 23
  23. 2024Generalizing Denoising to Non-Equilibrium Structures Improves Equivariant Force Fieldscited by 18
  24. 2025Towards Fast, Specialized Machine Learning Force Fields: Distilling Foundation Models via Energy Hessianscited by 15
  25. 2024Probing the effects of broken symmetries in machine learningcited by 13
  26. 2024Predicting solvation free energies with an implicit solvent machine learning potentialcited by 10
  27. 2024Nutmeg and SPICE: Models and Data for Biomolecular Machine Learningcited by 7
  28. 2024Hessian QM9: A quantum chemistry database of molecular Hessians in implicit solventscited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 37; 2026: 42.

79 papers.

Sort:

By area:

  1. 2025UMA: A Family of Universal Models for Atomscited by 99
  2. 2025Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matchingcited by 78
  3. 2025Adjoint Schr"odinger Bridge Samplercited by 20
  4. 2025Mol-LLaMA: Towards General Understanding of Molecules in Large Molecular Language Modelcited by 16
  5. 2025Space Group Equivariant Crystal Diffusioncited by 15
  6. 2025Universally Converging Representations of Matter Across Scientific Foundation Modelscited by 12
  7. 2025Reinforcement Fine-Tuning for Materials Designcited by 11
  8. 2025PLaID++: A Preference Aligned Language Model for Targeted Inorganic Materials Designcited by 11
  9. 2025HIP: Hessian Interatomic Potentials without derivativescited by 11
  10. 2026Pushing the limits of unconstrained machine-learned interatomic potentialscited by 11
  11. 2025How simple can you go? An off-the-shelf transformer approach to molecular dynamicscited by 10
  12. 2025A universal augmentation framework for long-range electrostatics in machine learning interatomic potentialscited by 9
38 more
  1. 2025MOFSimBench: Evaluating Universal Machine Learning Interatomic Potentials In Metal--Organic Framework Molecular Modelingcited by 7
  2. 2025The Open DAC 2025 Dataset for Sorbent Discovery in Direct Air Capturecited by 7
  3. 2025Learning from the electronic structure of molecules across the periodic tablecited by 7
  4. 2026MACE-POLAR-1: A Polarisable Electrostatic Foundation Model for Molecular Chemistrycited by 7
  5. 2026A recipe for scalable attention-based MLIPs: unlocking long-range accuracy with all-to-all node attentioncited by 7
  6. 2026EquiformerV3: Scaling Efficient, Expressive, and General SE(3)-Equivariant Graph Attention Transformerscited by 7
  7. 2025chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom MD simulationscited by 6
  8. 2025BLIPs: Bayesian Learned Interatomic Potentialscited by 6
  9. 2025Training a Foundation Model for Materials on a Budgetcited by 6
  10. 2025Transformers Discover Molecular Structure Without Graph Priorscited by 6
  11. 2026MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interatomic Potentialscited by 6
  12. 2025UniFFBench: Evaluating Universal Machine Learning Force Fields Against Experimental Measurementscited by 5
  13. 2025Spectral/Spatial Tensor Atomic Cluster Expansion with Universal Embeddings in Cartesian Spacecited by 5
  14. 2025Learning Inter-Atomic Potentials without Explicit Equivariancecited by 5
  15. 2026PFT: Phonon Fine-tuning for Machine Learned Interatomic Potentialscited by 5
  16. 2025DistMLIP: A Distributed Inference Platform for Machine Learning Interatomic Potentialscited by 4
  17. 2025Scaling Laws and Symmetry, Evidence from Neural Force Fieldscited by 4
  18. 2025AQCat25: Unlocking spin-aware, high-fidelity machine learning potentials for heterogeneous catalysiscited by 4
  19. 2026Flow Sampling: Learning to Sample from Unnormalized Densities via Denoising Conditional Processescited by 4
  20. 2025Understanding and Mitigating Distribution Shifts For Machine Learning Force Fieldscited by 3
  21. 2025AI-Driven Expansion and Application of the Alexandria Databasecited by 3
  22. 2025PhononBench:A Large-Scale Phonon-Based Benchmark for Dynamical Stability in Crystal Generationcited by 3
  23. 2026E2Former-V2: On-the-Fly Equivariant Attention with Linear Activation Memorycited by 3
  24. 2026Elign: Equivariant Diffusion Model Alignment from Foundational Machine Learning Force Fieldscited by 3
  25. 2026Zatom-1: Towards a Multimodal Foundation Model for 3D Molecules and Materialscited by 3
  26. 2025Polymorphism Crystal Structure Prediction with Adaptive Space Group Diversity Controlcited by 2
  27. 2025Advancing Universal Deep Learning for Electronic-Structure Hamiltonian Prediction of Materialscited by 2
  28. 2025A Cartesian-3j Framework for Machine Learning Interatomic Potentialscited by 2
  29. 2026Approximate Equivariance via Projection-based Regularisationcited by 2
  30. 2026Better without U: Impact of Selective Hubbard U Correction on Foundational MLIPscited by 2
  31. 2026Smooth Dynamic Cutoffs for Machine Learning Interatomic Potentialscited by 2
  32. 2026CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Modelscited by 2
  33. 2025Tensor Decomposition Networks for Fast Machine Learning Interatomic Potential Computationscited by 1
  34. 2025Gradient-based grand canonical optimization enabled by graph neural networks with fractional atomic existencecited by 1
  35. 2025Efficient Nudged Elastic Band Method using Neural Network Bayesian Algorithm Executioncited by 1
  36. 2026MEIDNet: Multimodal generative AI framework for inverse materials designcited by 1
  37. 2026From Evaluation to Design: Using Potential Energy Surface Smoothness Metrics to Guide Machine Learning Interatomic Potential Architecturescited by 1
  38. 2026Mixture of experts architectures for machine learning interatomic potentialscited by 1

50 most cited of 79 All 79 (JSON)