JanusPipe introduces SymFold and WaveK to enable efficient 3D-parallel training for conservative MLIPs, reporting 1.51x and 1.45x average throughput gains over 1F1B and Hanayo baselines on 32 GPUs.
A Graph Neural Network for the Era of Large Atomistic Models
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
fields
cond-mat.mtrl-sci 5 cs.DC 2 cond-mat.dis-nn 1 cs.LG 1 physics.chem-ph 1 physics.comp-ph 1 physics.geo-ph 1roles
background 2polarities
background 2representative citing papers
A co-trained multifidelity mixture-of-experts MLIP partitions simulations into high- and low-capacity regions, maintains exact energy conservation and bulk modulus alignment, and runs more than twice as fast as a single high-fidelity model on a Pt+CO system.
A deep free energy model learns the SCHA free energy surface to enable high-throughput crystal structure prediction with finite-temperature and nuclear quantum effects, reproducing known La-Sc-H phases and discovering a new stable LaScH8 structure.
A neural-network machine learning interatomic potential is used to estimate eutectic points in high-melting alloys by operating directly in the liquid phase without solid-structure input, demonstrated on the Ti-B-C system.
CrystalREPA closes the representation gap between crystal generators and universal MLIPs via contrastive alignment, yielding more stable and valid generated crystals while revealing that MLIP teacher quality is better predicted by representation distinguishability than by leaderboard accuracy.
Structural pruning of SO(3) equivariant atomistic models from large checkpoints yields 1.5-4x fewer parameters and 2.5-4x less pre-training compute than small models trained from scratch, while outperforming them on most Matbench Discovery metrics and downstream tasks.
CarNet is an equivariant graph-neural-network framework built on irreducible Cartesian natural tensors that predicts interatomic potentials and high-rank tensorial properties such as the elastic constant tensor.
An AI-driven workflow using Wyckoff substitution, MatterGen generation, DPA3 ML potentials, and DFT identifies 3574 new stable phosphides including 196 semiconductors and screens 30 optoelectronic plus 26 thermoelectric candidates.
GROMACS now runs multi-GPU DeePMD inference for molecular dynamics, reaching 40-66% strong scaling efficiency up to 32 devices on a 15k-atom protein system with over 90% time in inference.
Different uMLIPs encode chemical space in distinct ways, with high cross-model feature reconstruction errors, and fine-tuning preserves strong pre-training bias in the latent features.
Foundational atomistic models reproduce some structural and dynamical properties of iron alloys under core conditions but none consistently match first-principles benchmarks due to missing explicit treatment of thermal electronic excitations.
SLUSCHI-UP deploys the SLUSCHI melting-temperature workflow as a web service backed by selectable universal ML interatomic potentials with reported validation errors on benchmark sets.
citing papers explorer
-
JanusPipe: Efficient Pipeline Parallel Training for Machine Learning Interatomic Potentials
JanusPipe introduces SymFold and WaveK to enable efficient 3D-parallel training for conservative MLIPs, reporting 1.51x and 1.45x average throughput gains over 1F1B and Hanayo baselines on 32 GPUs.
-
Mixture of Experts Framework in Machine Learning Interatomic Potentials for Atomistic Simulations
A co-trained multifidelity mixture-of-experts MLIP partitions simulations into high- and low-capacity regions, maintains exact energy conservation and bulk modulus alignment, and runs more than twice as fast as a single high-fidelity model on a Pt+CO system.
-
Crystal structure prediction with nuclear quantum and finite-temperature effects via deep free energy learning
A deep free energy model learns the SCHA free energy surface to enable high-throughput crystal structure prediction with finite-temperature and nuclear quantum effects, reproducing known La-Sc-H phases and discovering a new stable LaScH8 structure.
-
Machine learning potential as a guide for eutectic in ultra-refractory multicomponent ceramics
A neural-network machine learning interatomic potential is used to estimate eutectic points in high-melting alloys by operating directly in the liquid phase without solid-structure input, demonstrated on the Ti-B-C system.
-
CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models
CrystalREPA closes the representation gap between crystal generators and universal MLIPs via contrastive alignment, yielding more stable and valid generated crystals while revealing that MLIP teacher quality is better predicted by representation distinguishability than by leaderboard accuracy.
-
Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning
Structural pruning of SO(3) equivariant atomistic models from large checkpoints yields 1.5-4x fewer parameters and 2.5-4x less pre-training compute than small models trained from scratch, while outperforming them on most Matbench Discovery metrics and downstream tasks.
-
Atomistic Machine Learning with Irreducible Cartesian Natural Tensors
CarNet is an equivariant graph-neural-network framework built on irreducible Cartesian natural tensors that predicts interatomic potentials and high-rank tensorial properties such as the elastic constant tensor.
-
Robust AI-Driven Discovery of Electronic Metal Phosphide Semiconductors
An AI-driven workflow using Wyckoff substitution, MatterGen generation, DPA3 ML potentials, and DFT identifies 3574 new stable phosphides including 196 semiconductors and screens 30 optoelectronic plus 26 thermoelectric candidates.
-
Making Room for AI: Multi-GPU Molecular Dynamics with Deep Potentials in GROMACS
GROMACS now runs multi-GPU DeePMD inference for molecular dynamics, reaching 40-66% strong scaling efficiency up to 32 devices on a 15k-atom protein system with over 90% time in inference.
-
Comparing the latent features of universal machine-learning interatomic potentials
Different uMLIPs encode chemical space in distinct ways, with high cross-model feature reconstruction errors, and fine-tuning preserves strong pre-training bias in the latent features.
-
Assessing foundational atomistic models for iron alloys under Earth's core conditions
Foundational atomistic models reproduce some structural and dynamical properties of iron alloys under core conditions but none consistently match first-principles benchmarks due to missing explicit treatment of thermal electronic excitations.
-
SLUSCHI-UP: A Web Infrastructure for SLUSCHI Melting-Temperature Calculations Using Universal Machine-Learning Interatomic Potentials
SLUSCHI-UP deploys the SLUSCHI melting-temperature workflow as a web service backed by selectable universal ML interatomic potentials with reported validation errors on benchmark sets.