The paper introduces prototype-based latent alignment to enable data-efficient fine-tuning of ML interatomic potentials, achieving up to 18% lower energy MAE than standard fine-tuning on rMD17 and SPICE benchmarks in low-data settings.
Canonical reference
Bartók, Mike C
Canonical reference. 100% of citing Pith papers cite this work as background.
10
Pith papers citing it
3,143
external citations · OpenAlex
Background
100% of classified citations
citation-role summary
background 5
citation-polarity summary
years
2026 10roles
background 5polarities
background 5representative citing papers
CUTS-GPR performs numerically exact Gaussian process regression with near-linear scaling in training points N and low-order polynomial scaling in dimensions D by exploiting additive kernels on incomplete grids.
Molecular dynamics simulations find that both I and MA defects in MAPbI3 diffuse rapidly at room temperature with barriers of 0.15-0.20 eV, with MA interstitials moving via concerted mechanisms and no MA vacancy migration observed.
A 1.62-trillion-atom molecular dynamics simulation achieves ab initio accuracy with 100x speedup over prior machine learning force fields and 86.9% weak scaling to 45,000 GPGPUs.
QCOF ML potentials tuned on COF data outperform general MACE models for defective systems and reveal higher thermal defect sensitivity in CTF-1 versus COF-LZU1 with nearly invariant low-strain mechanics.
A pretrained universal MLIP without spin or Hubbard-U corrections predicts the same chemisorbed S removal and O uptake under 15 eV O+/O2+ bombardment of WS2 as after three rounds of PBE+D3+U+spin fine-tuning, which still reduces energy/force MAE to 4.5 meV/atom and 0.076 eV/A.
GRACE-FS is generally more accurate and transferable while UNEP-v1 is much faster and supports uncertainty-aware multi-million-atom shock simulations of multicomponent alloys.
This perspective article develops a definition of foundational MLIPs and poses six open questions that the authors believe will define future research in machine-learned interatomic potentials.
citing papers explorer
-
Prototype-Guided Latent Alignment for Data-Efficient Fine-Tuning of Molecular Foundation Models
The paper introduces prototype-based latent alignment to enable data-efficient fine-tuning of ML interatomic potentials, achieving up to 18% lower energy MAE than standard fine-tuning on rMD17 and SPICE benchmarks in low-data settings.
-
Don't Get Your Kroneckers in a Twist: Gaussian Processes on High-Dimensional Incomplete Grids
CUTS-GPR performs numerically exact Gaussian process regression with near-linear scaling in training points N and low-order polynomial scaling in dimensions D by exploiting additive kernels on incomplete grids.
-
A Unified microscopic picture of cation and anion migration in MAPbI$_3$
Molecular dynamics simulations find that both I and MA defects in MAPbI3 diffuse rapidly at room temperature with barriers of 0.15-0.20 eV, with MA interstitials moving via concerted mechanisms and no MA vacancy migration observed.
-
Trillion-atom molecular dynamics simulations with ab initio accuracy
A 1.62-trillion-atom molecular dynamics simulation achieves ab initio accuracy with 100x speedup over prior machine learning force fields and 86.9% weak scaling to 45,000 GPGPUs.
-
Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks
QCOF ML potentials tuned on COF data outperform general MACE models for defective systems and reveal higher thermal defect sensitivity in CTF-1 versus COF-LZU1 with nearly invariant low-strain mechanics.
-
Fine-Tuning a Universal Machine-Learned Interatomic Potential for Oxygen Plasma Interactions with WS$_2$
A pretrained universal MLIP without spin or Hubbard-U corrections predicts the same chemisorbed S removal and O uptake under 15 eV O+/O2+ bombardment of WS2 as after three rounds of PBE+D3+U+spin fine-tuning, which still reduces energy/force MAE to 4.5 meV/atom and 0.076 eV/A.
-
Benchmarking Chemically Scalable Machine-Learning Interatomic Potentials for Large-Scale Simulations of Multicomponent Alloys
GRACE-FS is generally more accurate and transferable while UNEP-v1 is much faster and supports uncertainty-aware multi-million-atom shock simulations of multicomponent alloys.
-
Six Open Questions in Machine-Learned Interatomic Potential Foundation Models
This perspective article develops a definition of foundational MLIPs and poses six open questions that the authors believe will define future research in machine-learned interatomic potentials.
- Economised path integrals
- Spatial statistics for screening molecular structures