Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2602.16908

Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials

G. Laskaris, D. Morozov, D. Tarpanov, A. Seth, J. Procelewska, G. Sai Gautam, A. Sagingalieva, R. Brasher, A. Melnikov

cond-mat.mtrl-scics.LGquant-phSubmitted Feb 18, 2026• Announced Feb 20, 2026• Revised Jun 4, 2026 (v2)
v1 · Feb 18, 2026v2 · Jun 4, 2026
arXivHTMLPDF

Abstract

Allegro is a machine learning interatomic potential model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason, we apply multi-objective hyperparameter optimization to both objectives. Additionally, we experiment with modified architectures by constructing variants of Allegro: one extended with additional classical layers and one incorporating quantum-classical hybrid layers. We evaluate all models on QM9, rMD17-aspirin, rMD17-benzene, and a self-generated dataset of copper-lithium structures. As results, both variants surpass Allegro in force prediction accuracy across multiple datasets. The classical variant consistently improves over the baseline, while the quantum-classical hybrid variant achieves the best overall force prediction accuracy on the Cu-Li dataset, where it was fully optimized, outperforming the classical variant by approximately 13%. Notably, the hybrid variant also achieves competitive results on the remaining datasets despite using hyperparameters transferred from Cu-Li without dataset-specific optimization, suggesting that quantum-classical hybridization is a promising direction for enhancing MLIP architectures.

Journal reference
Comput. Mater. Sci. 270, 114742 (2026)
DOI
10.1016/j.commatsci.2026.114742
Comments
15 pages, 7 figures, 6 tables

Related papers

Towards Improved Quantum Machine Learning for Molecular Force Fields

Yannick Couzinié, Daimon Shunsuke, Hirofumi Nishi +3 more

This study explores the use of equivariant quantum neural networks (QNN) for generating molecular force fields, focusing on the rMD17 dataset. We consider a QNN architecture based on previous research and point out shortcomings in the parametrization of the atomic environments, that limits its expressivity as an interatomic potential and precludes transferability between molecules. We propose a revised QNN architecture that addresses these shortcomings. While both QNNs show promise in force prediction, with the revised architecture showing improved accuracy, they struggle with energy prediction. Further, both QNNs architectures fail to demonstrate a meaningful scaling law of decreasing errors with increasing training data. These findings highlight the challenges of scaling QNNs for complex molecular systems and emphasize the need for improved encoding strategies, regularization techniques, and hybrid quantum-classical approaches.

•Dec 1, 2025•0 citations

Quantum machine learning interatomic potential: Application of variational quantum algorithm

Kohei Numata, Wataru Mizukami, Kosuke Mitarai +2 more

This study applied quantum circuit learning, a commonly used hybrid quantum-classical machine learning algorithm, to a machine learning interatomic potential (MLIP) for predicting the energies of molecules in molecular datasets. We retrained the ANI model using the quantum transfer learning architecture [Mari et al., Quantum, 4:340, 2020] and evaluated numerical accuracy with a quantum circuit simulator. The evaluation confirmed that inserting a quantum circuit into the classical neural network of the MLIP yielded slightly higher accuracy than the fully classical neural network under certain conditions. In particular, the model incorporating a quantum circuit was more effective when the pretraining model had room for improvement in accuracy. These findings may contribute to advancing the application of quantum machine learning for MLIPs.

quant-pharXiv:2607.27841•Jul 31, 2026•0 citations

Towards improved quantum machine learning for molecular force fields

Yannick Couzinié, Shunsuke Daimon, Hirofumi Nishi +3 more

This study explores the use of equivariant quantum neural networks (QNN) for generating molecular force fields, focusing on the rMD17 dataset. We consider a QNN architecture based on previous research and point out shortcomings in the parametrization of the atomic environments. These shortcomings limit its expressivity as an interatomic potential and precludes transferability between molecules. We propose a revised QNN architecture that addresses these shortcomings. While both QNNs show promise in force prediction, with the revised architecture showing improved accuracy, they struggle with energy prediction. Further, both QNNs architectures fail to demonstrate a meaningful scaling law of decreasing errors with increasing training data. These findings highlight the challenges of scaling QNNs for complex molecular systems and emphasize the need for improved encoding strategies, regularization techniques, and hybrid quantum-classical approaches.

arXiv:2505.03213•Physical Review A•Nov 18, 2025•3 citations

A Hybrid Message-Passing and Electric-Field Representation for Electrostatic QM/MM Embedding with Machine Learning Interatomic Potentials

Raphael T. Husistein, Shobhit S. Chaturvedi, Thomas Weymuth +5 more

Within a hybrid quantum mechanics/molecular mechanics approach a small region of a large atomistic system is described quantum mechanically, while the rest of the system is represented by a classical force field. Despite this, the computational cost of the quantum calculation severely limits the accessible simulation and sampling time. Replacing the quantum calculation with a machine learning interatomic potential (MLIP) trained on quantum reference data can greatly increase the accessible simulation time. However, the accuracy and efficiency strongly depend on the specific way in which the MLIP incorporates information about the classical (electrostatic) environment. Here, we demonstrate that existing strategies based purely on features derived from an external electric field fail to distinguish certain environments, ultimately limiting their accuracy. As a solution, we propose an extension to the equivariant TensorNet architecture that mitigates this shortcoming by incorporating classical atoms close to the quantum region into the message-passing scheme, combined with features derived from the electric field for describing the environment farther away. We show that this combination achieves higher accuracy than either approach alone.

physics.chem-phCross-listedarXiv:2610.41523•Nov 2, 2026

A Critical Benchmark of Hybrid Quantum-Classical Graph Neural Networks for Molecular Property Prediction: Revealing Performance Bottlenecks in Current Architectures

Rakesh Kalshetty, Sreenivasa Manish, Shashank Kotagi +1 more

Hybrid quantum-classical neural networks (HQCNNs) represent a promising but largely unproven paradigm for enhancing molecular property prediction in drug discovery. Despite growing interest and optimistic claims in recent literature, rigorous benchmarks comparing these hybrid architectures against strong classical baselines remain scarce. This work implements and systematically evaluates a representative HQCNN architecture against its classical equivalent using standardized molecular datasets (QM9 and ZINC-250K) within a reproducible, multi-seed experimental framework. Contrary to optimistic expectations, our results reveal that the hybrid model consistently underperforms the classical baseline, showing a 49% increase in MAE on QM9 (1.0834 vs 0.7256) and a devastating 107% degradation on ZINC (1.4006 vs 0.6755), with R2values dropping by over 80% on both datasets. Through detailed analysis, we identify fundamental challenges including information bottlenecks during quantum encoding and difficult optimization landscapes in variational quantum circuits as primary causes of performance degradation. This work provides a crucial, datadriven reality check for the field, demonstrating that current “plug-and-play” quantum layer approaches may be insufficient for practical quantum advantages. All experimental code and data are publicly released to enable community replication and investigation of more sophisticated quantum-native architectures.

International Conference on Computation System and Information Technology for Sustainable Solutions (CSITSS)•Nov 20, 2025•0 citations

Parameter-Efficient Quantum Machine Learning for Multi-Class Brain Tumor Classification

Rahman Abdur, Tae-Hyong Kim

Medical image classification is a critical task in computer-aided diagnostics, but classical deep learning models often demand substantial computational resources and large parameter spaces. This paper investigates the potential of hybrid quantum-classical neural networks to address this challenge in the context of multi-class brain tumor classification from MRI scans. We introduce a custom hybrid quantum machine learning (QML) model and conduct a rigorous comparative analysis against its purely classical counterpart, as well as established architectures such as ResNet-18 and VGG-16. Our findings demonstrate that the proposed hybrid QML model achieves a state-of-the-art test accuracy of 96.58%, matching the performance of the significantly larger ResNet-18 model. Crucially, the hybrid model outperforms its direct classical counterpart, the Classical Benchmark (96.20%), while utilizing approximately 80% fewer parameters. Although simulating quantum circuits results in longer training times, our findings highlight the remarkable parameter efficiency and competitive accuracy of hybrid quantum models, suggesting a promising direction for developing powerful yet lightweight networks for medical diagnostics.

International Conference on Information and Communication Technology Convergence (ICTC)•Oct 14, 2025•1 citation
Allegro is a machine learning interatomic potential model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason, we apply multi-objective hyperparameter optimization to both objectives. Additionally, we experiment with modified architectures by constructing variants of Allegro: one extended with additional classical layers and one incorporating quantum-classical hybrid layers. We evaluate all models on QM9, rMD17-aspirin, rMD17-benzene, and a self-generated dataset of copper-lithium structures. As results, both variants surpass Allegro in force prediction accuracy across multiple datasets. The classical variant consistently improves over the baseline, while the quantum-classical hybrid variant achieves the best overall force prediction accuracy on the Cu-Li dataset, where it was fully optimized, outperforming the classical variant by approximately 13%. Notably, the hybrid variant also achieves competitive results on the remaining datasets despite using hyperparameters transferred from Cu-Li without dataset-specific optimization, suggesting that quantum-classical hybridization is a promising direction for enhancing MLIP architectures.
This study explores the use of equivariant quantum neural networks (QNN) for generating molecular force fields, focusing on the rMD17 dataset. We consider a QNN architecture based on previous research and point out shortcomings in the parametrization of the atomic environments, that limits its expressivity as an interatomic potential and precludes transferability between molecules. We propose a revised QNN architecture that addresses these shortcomings. While both QNNs show promise in force prediction, with the revised architecture showing improved accuracy, they struggle with energy prediction. Further, both QNNs architectures fail to demonstrate a meaningful scaling law of decreasing errors with increasing training data. These findings highlight the challenges of scaling QNNs for complex molecular systems and emphasize the need for improved encoding strategies, regularization techniques, and hybrid quantum-classical approaches.
This study applied quantum circuit learning, a commonly used hybrid quantum-classical machine learning algorithm, to a machine learning interatomic potential (MLIP) for predicting the energies of molecules in molecular datasets. We retrained the ANI model using the quantum transfer learning architecture [Mari et al., Quantum, 4:340, 2020] and evaluated numerical accuracy with a quantum circuit simulator. The evaluation confirmed that inserting a quantum circuit into the classical neural network of the MLIP yielded slightly higher accuracy than the fully classical neural network under certain conditions. In particular, the model incorporating a quantum circuit was more effective when the pretraining model had room for improvement in accuracy. These findings may contribute to advancing the application of quantum machine learning for MLIPs.
This study explores the use of equivariant quantum neural networks (QNN) for generating molecular force fields, focusing on the rMD17 dataset. We consider a QNN architecture based on previous research and point out shortcomings in the parametrization of the atomic environments. These shortcomings limit its expressivity as an interatomic potential and precludes transferability between molecules. We propose a revised QNN architecture that addresses these shortcomings. While both QNNs show promise in force prediction, with the revised architecture showing improved accuracy, they struggle with energy prediction. Further, both QNNs architectures fail to demonstrate a meaningful scaling law of decreasing errors with increasing training data. These findings highlight the challenges of scaling QNNs for complex molecular systems and emphasize the need for improved encoding strategies, regularization techniques, and hybrid quantum-classical approaches.
Within a hybrid quantum mechanics/molecular mechanics approach a small region of a large atomistic system is described quantum mechanically, while the rest of the system is represented by a classical force field. Despite this, the computational cost of the quantum calculation severely limits the accessible simulation and sampling time. Replacing the quantum calculation with a machine learning interatomic potential (MLIP) trained on quantum reference data can greatly increase the accessible simulation time. However, the accuracy and efficiency strongly depend on the specific way in which the MLIP incorporates information about the classical (electrostatic) environment. Here, we demonstrate that existing strategies based purely on features derived from an external electric field fail to distinguish certain environments, ultimately limiting their accuracy. As a solution, we propose an extension to the equivariant TensorNet architecture that mitigates this shortcoming by incorporating classical atoms close to the quantum region into the message-passing scheme, combined with features derived from the electric field for describing the environment farther away. We show that this combination achieves higher accuracy than either approach alone.
Hybrid quantum-classical neural networks (HQCNNs) represent a promising but largely unproven paradigm for enhancing molecular property prediction in drug discovery. Despite growing interest and optimistic claims in recent literature, rigorous benchmarks comparing these hybrid architectures against strong classical baselines remain scarce. This work implements and systematically evaluates a representative HQCNN architecture against its classical equivalent using standardized molecular datasets (QM9 and ZINC-250K) within a reproducible, multi-seed experimental framework. Contrary to optimistic expectations, our results reveal that the hybrid model consistently underperforms the classical baseline, showing a 49% increase in MAE on QM9 (1.0834 vs 0.7256) and a devastating 107% degradation on ZINC (1.4006 vs 0.6755), with R2values dropping by over 80% on both datasets. Through detailed analysis, we identify fundamental challenges including information bottlenecks during quantum encoding and difficult optimization landscapes in variational quantum circuits as primary causes of performance degradation. This work provides a crucial, datadriven reality check for the field, demonstrating that current “plug-and-play” quantum layer approaches may be insufficient for practical quantum advantages. All experimental code and data are publicly released to enable community replication and investigation of more sophisticated quantum-native architectures.
Medical image classification is a critical task in computer-aided diagnostics, but classical deep learning models often demand substantial computational resources and large parameter spaces. This paper investigates the potential of hybrid quantum-classical neural networks to address this challenge in the context of multi-class brain tumor classification from MRI scans. We introduce a custom hybrid quantum machine learning (QML) model and conduct a rigorous comparative analysis against its purely classical counterpart, as well as established architectures such as ResNet-18 and VGG-16. Our findings demonstrate that the proposed hybrid QML model achieves a state-of-the-art test accuracy of 96.58%, matching the performance of the significantly larger ResNet-18 model. Crucially, the hybrid model outperforms its direct classical counterpart, the Classical Benchmark (96.20%), while utilizing approximately 80% fewer parameters. Although simulating quantum circuits results in longer training times, our findings highlight the remarkable parameter efficiency and competitive accuracy of hybrid quantum models, suggesting a promising direction for developing powerful yet lightweight networks for medical diagnostics.