Quantum Chemistry
MD17
Structure
1Node Feature Dim
Coverage
271Results
1Variants
49Papers
- Learning 3D Anisotropic Noise Distributions Improves Molecular Force Field Modeling Learning 3D Anisotropic Noise Distributions Improves Molecular Force Field Modeling
- 3D-GSRD: 3D Molecular Graph Auto-Encoder with Selective Re-mask Decoding 3D-GSRD: 3D Molecular Graph Auto-Encoder with Selective Re-mask Decoding
- ICLR 2025: PAINET: A Principled Efficient Transformer for 3D Dynamics Modeling PAINET: A Principled Efficient Transformer for 3D Dynamics Modeling ICLR 2025
- GeoRecon: Graph-Level Representation Learning for 3D Molecules via Reconstruction-Based Pretraining GeoRecon: Graph-Level Representation Learning for 3D Molecules via Reconstruction-Based Pretraining
- Efficient Prediction of SO(3)-Equivariant Hamiltonian Matrices via SO(2) Local Frames Efficient Prediction of SO(3)-Equivariant Hamiltonian Matrices via SO(2) Local Frames
- How simple can you go? An off-the-shelf transformer approach to molecular dynamics How simple can you go? An off-the-shelf transformer approach to molecular dynamics
- ICLR 2025: MolSpectra: Pre-training 3D Molecular Representation with Multi-modal Energy Spectra MolSpectra: Pre-training 3D Molecular Representation with Multi-modal Energy Spectra ICLR 2025
- ICLR 2025: Equivariant Masked position prediction for efficient molecular representation Equivariant Masked position prediction for efficient molecular representation ICLR 2025
- ICML 2025: Efficient and Scalable Density Functional Theory Hamiltonian Prediction through Adaptive Sparsity Efficient and Scalable Density Functional Theory Hamiltonian Prediction through Adaptive Sparsity ICML 2025
- Relaxed Equivariance via Multitask Learning Relaxed Equivariance via Multitask Learning
- NeurIPS 2024: Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? NeurIPS 2024
- Pre-training with fractional denoising to enhance molecular property prediction Pre-training with fractional denoising to enhance molecular property prediction
- ICLR 2024: FreeCG: Free the Design Space of Clebsch–Gordan Transform for Machine Learning Force Fields FreeCG: Free the Design Space of Clebsch–Gordan Transform for Machine Learning Force Fields ICLR 2024
- Molecule Graph Networks with Many-body Equivariant Interactions Molecule Graph Networks with Many-body Equivariant Interactions
- NeurIPS 2024: Infusing Self-Consistency into Density Functional Theory Hamiltonian Prediction via Deep Equilibrium Models Infusing Self-Consistency into Density Functional Theory Hamiltonian Prediction via Deep Equilibrium Models NeurIPS 2024
- TMLR 2024: SE3Set: Harnessing equivariant hypergraph neural networks for molecular representation learning SE3Set: Harnessing equivariant hypergraph neural networks for molecular representation learning TMLR 2024
- ICML 2024: UniCorn: A Unified Contrastive Learning Approach for Multi-view Molecular Representation Learning UniCorn: A Unified Contrastive Learning Approach for Multi-view Molecular Representation Learning ICML 2024
- NeurIPS 2024: SubGDiff: A Subgraph Diffusion Model to Improve Molecular Representation Learning SubGDiff: A Subgraph Diffusion Model to Improve Molecular Representation Learning NeurIPS 2024
- ICLR 2024: Clifford Group Equivariant Simplicial Message Passing Networks Clifford Group Equivariant Simplicial Message Passing Networks ICLR 2024
- Cartesian atomic cluster expansion for machine learning interatomic potentials Cartesian atomic cluster expansion for machine learning interatomic potentials
- ICLR 2024: On the Completeness of Invariant Geometric Deep Learning Models On the Completeness of Invariant Geometric Deep Learning Models ICLR 2024
- NeurIPS 2023: Molecule Joint Auto-Encoding: Trajectory Pretraining with 2D and 3D Diffusion Molecule Joint Auto-Encoding: Trajectory Pretraining with 2D and 3D Diffusion NeurIPS 2023
- ICLR 2023: Sliced Denoising: A Physics-Informed Molecular Pre-Training Method Sliced Denoising: A Physics-Informed Molecular Pre-Training Method ICLR 2023
- ICLR 2023: From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction ICLR 2023
- ICLR 2023: Fast, Expressive SE$(n)$ Equivariant Networks through Weight-Sharing in Position-Orientation Space Fast, Expressive SE$(n)$ Equivariant Networks through Weight-Sharing in Position-Orientation Space ICLR 2023
- Geometry-enhanced pretraining on interatomic potentials Geometry-enhanced pretraining on interatomic potentials
- NeurIPS 2023: QH9: A Quantum Hamiltonian Prediction Benchmark for QM9 Molecules QH9: A Quantum Hamiltonian Prediction Benchmark for QM9 Molecules NeurIPS 2023
- NeurIPS 2023: Symmetry-Informed Geometric Representation for Molecules, Proteins, and Crystalline Materials Symmetry-Informed Geometric Representation for Molecules, Proteins, and Crystalline Materials NeurIPS 2023
- NeurIPS 2023: TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials NeurIPS 2023
- ICML 2023: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining ICML 2023
- NeurIPS 2023: A new perspective on building efficient and expressive 3D equivariant graph neural networks A new perspective on building efficient and expressive 3D equivariant graph neural networks NeurIPS 2023
- NeurIPS 2023: Is Distance Matrix Enough for Geometric Deep Learning? Is Distance Matrix Enough for Geometric Deep Learning? NeurIPS 2023
- ICLR 2023: Spatial Attention Kinetic Network with E(n)-Equivariance Spatial Attention Kinetic Network with E(n)-Equivariance ICLR 2023
- ViSNet: an equivariant geometry-enhanced graph neural network with vector-scalar interactive message passing for molecules ViSNet: an equivariant geometry-enhanced graph neural network with vector-scalar interactive message passing for molecules
- TMLR 2022: Forces are not Enough: Benchmark and Critical Evaluation for Machine Learning Force Fields with Molecular Simulations Forces are not Enough: Benchmark and Critical Evaluation for Machine Learning Force Fields with Molecular Simulations TMLR 2022
- AAAI 2022: Energy-Motivated Equivariant Pretraining for 3D Molecular Graphs Energy-Motivated Equivariant Pretraining for 3D Molecular Graphs AAAI 2022
- ICLR 2022: Molecular Geometry Pretraining with SE(3)-Invariant Denoising Distance Matching Molecular Geometry Pretraining with SE(3)-Invariant Denoising Distance Matching ICLR 2022
- NeurIPS 2022: MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields NeurIPS 2022
- NeurIPS 2022: So3krates: Equivariant attention for interactions on arbitrary length-scales in molecular systems So3krates: Equivariant attention for interactions on arbitrary length-scales in molecular systems NeurIPS 2022
- The design space of E(3)-equivariant atom-centred interatomic potentials The design space of E(3)-equivariant atom-centred interatomic potentials
- TorchMD-NET: Equivariant Transformers for Neural Network based Molecular Potentials TorchMD-NET: Equivariant Transformers for Neural Network based Molecular Potentials
- NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces
- Rotation Invariant Graph Neural Networks using Spin Convolutions Rotation Invariant Graph Neural Networks using Spin Convolutions
- SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects
- ICLR 2021: Spherical Message Passing for 3D Molecular Graphs Spherical Message Passing for 3D Molecular Graphs ICLR 2021
- E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
- ICLR 2020: Directional Message Passing for Molecular Graphs Directional Message Passing for Molecular Graphs ICLR 2020
- PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges
- SchNet -- a deep learning architecture for molecules and materials SchNet -- a deep learning architecture for molecules and materials
Energy-and-force regression Ab-initio molecular dynamics — predict energy and forces for small organic molecules. Inputs are atomic species and molecular geometries. Evaluated by MAE.
Dataset details
Task
Ab-initio molecular dynamics — predict energy and forces for small organic molecules. The primary catalog metric is MAE.
Data
Ab-initio molecular dynamics — predict energy and forces for small organic molecules.
Features
Atoms, nuclear species, and molecular geometries are the core inputs. Graph edges or neighbor lists are model preprocessing choices; energy units, force signs, target normalization, and coordinate conventions must match the release.
Splits and evaluation
The catalog records these protocols or variants: Standard split. Evaluation uses MAE (lower is better). Exact masks or folds must come from the cited release.
Quirks and pitfalls
Near-duplicate conformations and molecule-wise versus conformation-wise splits can cause leakage. Units and reference-energy conventions differ across implementations, and MD trajectories are strongly autocorrelated.
Sources
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | MAE ↓ | ADE | E | Energy MAE (kcal mol⁻¹) | F | FDE | Force MAE (kcal mol⁻¹ Å⁻¹) | H | RMSD | epsilon | psi | Paper |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | QHNet | 1.04e-05 | — | — | — | — | — | — | — | — | — | — | |
| 2 | QHNetV2 | 1.04e-05 | — | — | — | — | — | — | — | — | — | — | |
| 3 | PhiSNet | 1.76e-05 | — | — | — | — | — | — | — | — | — | — | |
| 4 | SPHNet | 1.94e-05 | — | — | — | — | — | — | — | — | — | — | |
| 5 | QHNet | 8.31e-05 | — | — | — | — | — | — | — | — | — | — | |
| 6 | ViSNet | 0.0007 | — | — | — | — | — | — | — | — | — | — | |
| 7 | BOTNet | 0.0007 | — | — | — | — | — | — | — | — | — | — | |
| 8 | NequIP | 0.0009 | — | — | — | — | — | — | — | — | — | — | |
| 9 | GeoNGNN | 0.0014 | — | — | — | — | — | — | — | — | — | — | |
| 10 | UNiTE | 0.0020 | — | — | — | — | — | — | — | — | — | — | |
| 11 | PAINET | 0.0032
±
0.0001 RMSD
0.0018
±
8.00e-06 | — | — | — | — | — | — | — | 0.0018
±
8.00e-06 | — | — | |
| 12 | EGNO | 0.0037
±
0.0001 | — | — | — | — | — | — | — | — | — | — | |
| 13 | HEGNN | 0.0037
±
0.0002 | — | — | — | — | — | — | — | — | — | — | |
| 14 | GMN | 0.0040
±
0.0001 | — | — | — | — | — | — | — | — | — | — | |
| 15 | Allegro | 0.0046 | — | — | — | — | — | — | — | — | — | — | |
| 16 | EGNN | 0.0047
±
0.0002 | — | — | — | — | — | — | — | — | — | — | |
| 17 | rf | 0.0050
±
0.0001 | — | — | — | — | — | — | — | — | — | — | |
| 18 | GNN | 0.0054
±
1.00e-05 | — | — | — | — | — | — | — | — | — | — | |
| 19 | REMUL-GNN | 0.0054
±
1.00e-05 | — | — | — | — | — | — | — | — | — | — | |
| 20 | EGNNReg | 0.0063
±
0.0001 | — | — | — | — | — | — | — | — | — | — | |
| 21 | DA-GNN | 0.0069
±
1.00e-05 | — | — | — | — | — | — | — | — | — | — | |
| 22 | GemNet-T | 0.0070 Energy
0.6840 Force
0.5580 | — | — | 0.6840 | — | — | 0.5580 | — | — | — | — | |
| 23 | PaiNN | 0.0100 Energy
27.6260 Force
0.5720 | — | — | 27.6260 | — | — | 0.5720 | — | — | — | — | |
| 24 | Equiformer | 0.0200 Energy
0.3080 Force
0.2860 | — | — | 0.3080 | — | — | 0.2860 | — | — | — | — | |
| 25 | TensorNet 2L (770k) | 0.0200 | — | — | — | — | — | — | — | — | — | — | |
| 26 | GDML | 0.0200 | — | — | — | — | — | — | — | — | — | — | |
| 27 | PhysNet-ens5 | 0.0200 | — | — | — | — | — | — | — | — | — | — | |
| 28 | 3D-GSRD | 0.0266 | — | — | — | — | — | — | — | — | — | — | |
| 29 | Allegro | 0.0290 Energy
1.1380 Force
3.4050 | — | — | 1.1380 | — | — | 3.4050 | — | — | — | — | |
| 30 | BOTNet | 0.0300 E
2.3000 F
8.5000 | — | 2.3000 | — | 8.5000 | — | — | — | — | — | — | |
| 31 | TensorNet 1L (535k) | 0.0300 | — | — | — | — | — | — | — | — | — | — | |
| 32 | PhysNet | 0.0300 | — | — | — | — | — | — | — | — | — | — | |
| 33 | FreeCG | 0.0340 | — | — | — | — | — | — | — | — | — | — | |
| 34 | 3D-EMGP (TorchMD-NET) | 0.0352 | — | — | — | — | — | — | — | — | — | — | |
| 35 | AniDS | 0.0360 | — | — | — | — | — | — | — | — | — | — | |
| 36 | DeNS (Lmax=3) | 0.0370 | — | — | — | — | — | — | — | — | — | — | |
| 37 | TorchMD-Net | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 38 | NequIP | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 39 | QuinNet | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 40 | SE3Set | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 41 | ViSNet | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 42 | NequIP (Lmax=3) | 0.0390 | — | — | — | — | — | — | — | — | — | — | |
| 43 | NequIP | 0.0400 | — | — | — | — | — | — | — | — | — | — | |
| 44 | DTNN | 0.0400 | — | — | — | — | — | — | — | — | — | — | |
| 45 | 3D-ReTrans | 0.0417 | — | — | — | — | — | — | — | — | — | — | |
| 46 | ENINet | 0.0430 | — | — | — | — | — | — | — | — | — | — | |
| 47 | Frad | 0.0431 | — | — | — | — | — | — | — | — | — | — | |
| 48 | PhysNet-ens5 | 0.0440 | — | — | — | — | — | — | — | — | — | — | |
| 49 | UniCorn | 0.0460 | — | — | — | — | — | — | — | — | — | — | |
| 50 | Equiformer | 0.0460 | — | — | — | — | — | — | — | — | — | — |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time