Multitask learning with semiempirical orbital charges enables sample-efficient MLIPs

• URL: http://arxiv.org/abs/2605.24073v1
• Date: Fri, 22 May 2026 14:12:27 GMT
• Title: Multitask learning with semiempirical orbital charges enables sample-efficient MLIPs
• Abstract summary: We show that multitask learning utilizing orbitally resolved semiempirical charges significantly improves sample efficiency and accuracy in machine learning interatomic potentials.<n>Our model achieves a 46% reduction in energy mean absolute error and requires five times less data to match the performance of energy-only models.
• Score: 13.715688117301385
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Machine learning interatomic potentials (MLIPs) require generating computationally expensive, large-scale training datasets to accurately simulate materials and molecules. Incorporating electronic structure information using multitask learning improves sample efficiency, however, training on full Hamiltonian matrices, which scale quadratically with the number of atoms, is intractable for large datasets. In this work, we show that multitask learning utilizing orbitally resolved semiempirical charges significantly improves sample efficiency and accuracy in MLIPs. To efficiently predict orbital charges, we implement a specialized equivariant model, reducing charge prediction error compared to an invariant baseline. By augmenting training with computationally inexpensive GFN1-xTB orbital charges, which scale linearly with the number of atoms, our model achieves a 46\% reduction in energy mean absolute error and requires five times less data to match the performance of energy-only models. Furthermore, our approach outperforms models trained on expensive density functional theory (DFT) atomic charges, capturing orbitally resolved electronic complexity and forcing the network to learn a physically accurate latent space that spontaneously clusters metals by shared chemical properties. Because orbital charges are only required during training, this approach preserves inference efficiency, providing a scalable recipe for developing accurate, data-efficient foundation models for complex chemical systems.

Related papers

• Learning the Kohn-Sham map with neural operators for quasi-linear scaling density functional theory [57.90994363143731]
  Kohn--Sham density functional theory (DFT) underpins electronic-structure simulations.<n>Repeated orbital diagonalizations lead to cubic scaling, restricting quantum calculations to modest scales only.<n>We develop a learning target for orbital-free DFT that maps a Kohn--Sham potential directly to the corresponding density and noninteracting kinetic energy.
  arXiv  Detail & Related papers  (2026-08-24T23:05:42Z)
• Autotuning T-PaiNN: Enabling Data-Efficient GNN Interatomic Potential Development via Classical-to-Quantum Transfer Learning [0.0]
  We introduce a transfer learning framework, Transfer-PaiNN, that substantially improves the data efficiency of GNN-MLIPs by leveraging inexpensive classical force field data.<n>We demonstrate the effectiveness of autotuning T-PaiNN on both gas-phase molecular systems and condensed-phase liquid water.<n>For example, using the QM9 data set, error reductions of up to 25 times are observed in low-data regimes, while liquid water simulations show improved predictions of energies, forces, and experimentally relevant properties such as density and diffusion.
  arXiv  Detail & Related papers  (2026-03-25T19:22:56Z)
• Foundation Models for Discovery and Exploration in Chemical Space [57.97784111110166]
  MIST is a family of molecular foundation models trained on large unlabeled datasets.<n>We demonstrate the ability of these models to solve real-world problems across chemical space.
  arXiv  Detail & Related papers  (2025-10-20T17:56:01Z)
• Physical Consistency Bridges Heterogeneous Data in Molecular Multi-Task Learning [79.75718786477638]
  We exploit the specialty of molecular tasks that there are physical laws connecting them, and design consistency training approaches. We demonstrate that the more accurate energy data can improve the accuracy of structure prediction. We also find that consistency training can directly leverage force and off-equilibrium structure data to improve structure prediction.
  arXiv  Detail & Related papers  (2024-10-14T03:11:33Z)
• Gradual Optimization Learning for Conformational Energy Minimization [69.36925478047682]
  Gradual Optimization Learning Framework (GOLF) for energy minimization with neural networks significantly reduces the required additional data. Our results demonstrate that the neural network trained with GOLF performs on par with the oracle on a benchmark of diverse drug-like molecules.
  arXiv  Detail & Related papers  (2023-11-05T11:48:08Z)
• Electronic Structure Prediction of Multi-million Atom Systems Through Uncertainty Quantification Enabled Transfer Learning [5.4875371069660925]
  Ground state electron density -- obtainable using Kohn-Sham Density Functional Theory (KS-DFT) simulations -- contains a wealth of material information. However, the computational expense of KS-DFT scales cubically with system size which tends to stymie training data generation. Here, we address this fundamental challenge by employing transfer learning to leverage the multi-scale nature of the training data.
  arXiv  Detail & Related papers  (2023-08-24T21:41:29Z)
• Transfer learning for chemically accurate interatomic neural network potentials [0.0]
  We show that pre-training the network parameters on data obtained from density functional calculations improves the sample efficiency of models trained on more accurate ab-initio data. We provide GM-NN potentials pre-trained and fine-tuned on the ANI-1x and ANI-1ccx data sets, which can easily be fine-tuned on and applied to organic molecules.
  arXiv  Detail & Related papers  (2022-12-07T19:21:01Z)
• Electronic-structure properties from atom-centered predictions of the electron density [0.0]
  electron density of a molecule or material has recently received major attention as a target quantity of machine-learning models. We propose a gradient-based approach to minimize the loss function of the regression problem in an optimized and highly sparse feature space. We show that starting from the predicted density a single Kohn-Sham diagonalization step can be performed to access total energy components that carry an error of just 0.1 meV/atom.
  arXiv  Detail & Related papers  (2022-06-28T15:35:55Z)
• BIGDML: Towards Exact Machine Learning Force Fields for Materials [55.944221055171276]
  Machine-learning force fields (MLFF) should be accurate, computationally and data efficient, and applicable to molecules, materials, and interfaces thereof. Here, we introduce the Bravais-Inspired Gradient-Domain Machine Learning approach and demonstrate its ability to construct reliable force fields using a training set with just 10-200 atoms.
  arXiv  Detail & Related papers  (2021-06-08T10:14:57Z)
• Multi-task learning for electronic structure to predict and explore molecular potential energy surfaces [39.228041052681526]
  We refine the OrbNet model to accurately predict energy, forces, and other response properties for molecules. The model is end-to-end differentiable due to the derivation of analytic gradients for all electronic structure terms. It is shown to be transferable across chemical space due to the use of domain-specific features.
  arXiv  Detail & Related papers  (2020-11-05T06:48:46Z)
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features [42.96944345045462]
  textscOrbNet is shown to outperform existing methods in terms of learning efficiency and transferability. For applications to datasets of drug-like molecules, textscOrbNet predicts energies within chemical accuracy of DFT at a computational cost that is thousand-fold or more reduced.
  arXiv  Detail & Related papers  (2020-07-15T22:38:41Z)
• Graph Neural Network for Hamiltonian-Based Material Property Prediction [56.94118357003096]
  We present and compare several different graph convolution networks that are able to predict the band gap for inorganic materials. The models are developed to incorporate two different features: the information of each orbital itself and the interaction between each other. The results show that our model can get a promising prediction accuracy with cross-validation.
  arXiv  Detail & Related papers  (2020-05-27T13:32:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.