发表机构
Technical University of Munich (TUM); Munich Data Science Institute (MDSI); Munich Center for Machine Learning (MCML); NVIDIA（慕尼黑工业大学; 慕尼黑数据科学研究所; 慕尼黑机器学习中心; 英伟达）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出物理对齐的电子基态描述符模型，通过施加物理约束和优化架构，在尺寸外推和反应化学任务中显著提升泛化能力，大幅降低能量和力误差。
AI 中文摘要
机器学习原子间势（MLIPs）在分布内任务中表现出色，加速了药物和材料开发，但它们在分布外泛化方面存在困难。我们提出通过设计可观测无关的电子基态描述符模型（GSMs）来推动成本-精度帕累托前沿，其计算成本介于MLIPs和Kohn-Sham密度泛函理论（KS-DFT）之间。我们通过施加物理约束并移除对非物理或无关自由度的优化压力，将GSMs的学习目标和架构与KS-DFT的控制方程对齐。在我们从QM9到QM40的尺寸外推实验中，我们的组合贡献OrthoNormal-Loss（ON-Loss）和Grassmann Restricted Occupied-Orbital Training（GROOT）相比之前最先进的密度GSMs，能量和力平均绝对误差（MAE）分别降低了79.1%和83.4%。对于哈密顿量GSMs，ON-Loss和Residual Optimal-gauge Conditioning-aware KS-Eq. Training（ROCKET）共同将最强基线的能量和力MAE分别降低了99.8%和95.9%。使用自洽性拒绝准则，我们在QMugs上过滤外推误差，拒绝了不到0.4%的预测，同时达到0.07 mHa的能量MAE。最后，我们展示了无标签自洽性微调的效率，并将GSMs迁移到Transition1x中的反应化学，达到低于化学精度的能量误差。
英文摘要
Machine-learned interatomic potentials (MLIPs) excel at in-distribution tasks, accelerating drug and material development, yet they struggle to generalize out-of-distribution. We propose to push the cost-accuracy Pareto frontier by designing observable-agnostic electronic ground-state descriptor models (GSMs) with computational costs situated between MLIPs and Kohn-Sham density functional theory (KS-DFT). We align the learning objectives and architectures of GSMs with the governing equations of KS-DFT by enforcing physical constraints and removing optimization pressure on unphysical or irrelevant degrees of freedom. In our size-extrapolation experiments from QM9 to QM40, our combined contributions OrthoNormal-Loss (ON-Loss) and Grassmann Restricted Occupied-Orbital Training (GROOT) reach a 79.1% energy and 83.4% force mean absolute error (MAE) reduction over previous state-of-the-art density GSMs. For Hamiltonian GSMs, ON-Loss and Residual Optimal-gauge Conditioning-aware KS-Eq. Training (ROCKET) together reduce the energy and force MAEs of the strongest baseline by 99.8% and 95.9%, respectively. Using a self-consistency rejection criterion, we filter out extrapolation errors on QMugs, rejecting fewer than 0.4% of predictions while reaching an energy MAE of 0.07 mHa. Finally, we demonstrate the efficiency of label-free self-consistency fine-tuning, and transfer GSMs to reactive chemistry in Transition1x, reaching energy errors below chemical accuracy.
