Scaling Machine Learning Interatomic Potentials with Mixtures of Experts
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912962846916608 |
|---|---|
| author | Liu, Yuzhi Zhang, Duo Peng, Anyang E, Weinan Zhang, Linfeng Wang, Han |
| author_facet | Liu, Yuzhi Zhang, Duo Peng, Anyang E, Weinan Zhang, Linfeng Wang, Han |
| contents | Machine Learning Interatomic Potentials (MLIPs) enable accurate large-scale atomistic simulations, yet improving their expressive capacity efficiently remains challenging. Here we systematically develop Mixture-of-Experts (MoE) and Mixture-of-Linear-Experts (MoLE) architectures for MLIPs and analyze the effects of routing strategies and expert designs. We show that sparse activation combined with shared experts yields substantial performance gains, and that nonlinear MoE formulations outperform MoLE when shared experts are present, underscoring the importance of nonlinear expert specialization. Furthermore, element-wise routing consistently surpasses configuration-level routing, while global MoE routing often leads to numerical instability. The resulting element-wise MoE model achieves state-of-the-art accuracy across the OMol25, OMat24, and OC20M benchmarks. Analysis of routing patterns reveals chemically interpretable expert specialization aligned with periodic-table trends, indicating that the model effectively captures element-specific chemical characteristics for precise interatomic modeling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_07977 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Scaling Machine Learning Interatomic Potentials with Mixtures of Experts Liu, Yuzhi Zhang, Duo Peng, Anyang E, Weinan Zhang, Linfeng Wang, Han Chemical Physics Machine Learning Computational Physics Machine Learning Interatomic Potentials (MLIPs) enable accurate large-scale atomistic simulations, yet improving their expressive capacity efficiently remains challenging. Here we systematically develop Mixture-of-Experts (MoE) and Mixture-of-Linear-Experts (MoLE) architectures for MLIPs and analyze the effects of routing strategies and expert designs. We show that sparse activation combined with shared experts yields substantial performance gains, and that nonlinear MoE formulations outperform MoLE when shared experts are present, underscoring the importance of nonlinear expert specialization. Furthermore, element-wise routing consistently surpasses configuration-level routing, while global MoE routing often leads to numerical instability. The resulting element-wise MoE model achieves state-of-the-art accuracy across the OMol25, OMat24, and OC20M benchmarks. Analysis of routing patterns reveals chemically interpretable expert specialization aligned with periodic-table trends, indicating that the model effectively captures element-specific chemical characteristics for precise interatomic modeling. |
| title | Scaling Machine Learning Interatomic Potentials with Mixtures of Experts |
| topic | Chemical Physics Machine Learning Computational Physics |
| url | https://arxiv.org/abs/2603.07977 |