Research map / Probabilistic and scientific modeling / Molecules and proteins
Machine-learned interatomic potentials
Learns and benchmarks interatomic potentials for atomistic simulation of materials.
74 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Molecules and proteins, keyword cluster “molecules · drug · chemical”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
15 papers in 2023–24 and 48 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 3.1% to 5.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · SchNet: A continuous-filter convolutional neural network for modeling quantum interactions (NeurIPS 2017 · 470 citations)
- 2018 · End-to-end Symmetry Preserving Inter-atomic Potential Energy Model for Finite and Extended Systems (NeurIPS 2018 · 218 citations)
- 2019 · Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective (AAAI 2019 · 163 citations)
- 2021 · Non-Autoregressive Electron Redistribution Modeling for Reaction Prediction (ICML 2021 · 11 citations)
- 2022 · Crystal Diffusion Variational Autoencoder for Periodic Material Generation (ICLR 2022 · 105 citations)
- 2023 · Sampling-free Inference for Ab-Initio Potential Energy Surface Networks (ICLR 2023 · 4 citations)
- 2024 · A Recipe for Charge Density Prediction (NeurIPS 2024 · 4 citations)
- 2025 · Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction (ICML 2025 · 33 citations)
- 2026 · Revisiting the Canonicalization for Fast and Accurate Crystal Tensor Property Prediction (AAAI 2026 · 3 citations)
Earliest papers here
- SchNet: A continuous-filter convolutional neural network for modeling quantum interactions (NeurIPS 2017 · 470 citations)
- End-to-end Symmetry Preserving Inter-atomic Potential Energy Model for Finite and Extended Systems (NeurIPS 2018 · 218 citations)
- Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective (AAAI 2019 · 163 citations)
What has been tried
40 papers here report 414 measured changes (a component added, removed or swapped, compared with a baseline): helped 61%, no clear effect 9%, hurt 25%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use conservative forces instead of direct-force prediction | SPICE-MACE-OFF · Energy MAE | 0.56 -> 0.23 | Helped | Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction (ICML 2025 · 33 citations) |
| Pretrain ChemBERTa with masked language modeling | BOOM OOD tasks · OOD RMSE | negligible change | No clear effect | BOOM: Benchmarking Out-Of-distribution Molecular Property Predictions of Machine Learning Models (NeurIPS 2025 · 9 citations) |
| Add an energy head and energy targets to the direct-force PET model | Bulk water test set · Force MAE | 24.3 -> 24.8 | No clear effect | The dark side of the forces: assessing non-conservative force models for atomistic machine learning (ICML 2025 · 4 citations) |
| increase eSCN model size from K=4, L=3 to K=8, L=6 | QM9 test · NMAE [%] | 0.504% to 0.434% | Helped | A Recipe for Charge Density Prediction (NeurIPS 2024 · 4 citations) |
| replace full envelopes with memory-efficient envelopes | molecular systems tested · performance | Helped | Neural Pfaffians: Solving Many Many-Electron Schrödinger Equations (NeurIPS 2024 · 2 citations) | |
| Use an Arrhenius physics-based predictive head instead of a regular fully connected head | IlThermo ln(κ), temperature-range exclusion · MAE | 0.354 ± 0.152 -> 0.284 ± 0.025 | Helped | CheMixHub: Datasets and Benchmarks for Chemical Mixture Property Prediction (NeurIPS 2025 · 1 citation) |
| add lattice angle and atom-count N information to CSPNet | carbon-NXL (all N) · standard match rate | 47.3\,\% -> 47.7\,\% | Helped | All that structure matches does not glitter (NeurIPS 2025 · 1 citation) |
| add pairwise interaction modeling to self-attention | BH reactant-based and SM ligand-based out-of-sample · R² | 0.430 ± 0.038 -> 0.480 ± 0.032; 0.506 ± 0.026 -> 0.549 ± 0.019 | Helped | Reaction Prediction via Interaction Modeling of Symmetric Difference Shingle Sets (NeurIPS 2025 · 0 citations) |
| Add negative Gaussian weights | QM9 test · NMAE [%] | 2.027 -> 0.176 | Helped | ELECTRA: A Cartesian Network for 3D Charge Density Prediction with Floating Orbitals (NeurIPS 2025 · 0 citations) |
| Add exponential distance-dependent spring coefficients rather than inverse-square coefficients | QM9 · Molecule stability (%) | 84.6 ± 1.8 -> 89.8 ± 2.8 | Helped | Energy Loss Functions for Physical Systems (NeurIPS 2025 · 0 citations) |
| Raise the ED threshold from ρτ=0.1 to ρτ=0.2 | ED5-EDP · MAE | 0.3362±0.2900 -> 0.0448±0.0133 | Helped | EDBench: Large-Scale Electron Density Data for Molecular Modeling (NeurIPS 2025 · 0 citations) |
| Add Group CG transform | MD17 Aspirin · Force MAE | 0.155 -> 0.144 | Helped | FreeCG: Free the Design Space of Clebsch-Gordan Transform for Machine Learning Force Fields (ICLR 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 24 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- QM9 · 5 of 24 papers
- MD22 · 4 of 24 papers
- SPICE · 4 of 24 papers
- MPTrj · 3 of 24 papers
- MP-20 · 3 of 24 papers
- rMD17 · 2 of 24 papers
- Alex-MP-20 · 2 of 24 papers
- PCQM4Mv2 · 2 of 24 papers
Baselines
- EquiformerV2 · 4 of 24 papers
- MACE · 4 of 24 papers
- DimeNet++ · 4 of 24 papers
- ORB-v2 · 3 of 24 papers
- MACE-MP-0 · 3 of 24 papers
- SchNet · 3 of 24 papers
- PaiNN · 3 of 24 papers
- SevenNet · 2 of 24 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 63 papers; from OpenAlex and the papers' affiliations).
- Massachusetts Institute of Technology · 4 papers
- University of California, Berkeley · 4 papers
- Technical University Munich · 3 papers
- Microsoft · 3 papers
- Shanghai Jiao Tong University · 2 papers
- Georgia Institute of Technology · 2 papers
- ByteDance · 2 papers
- EPFL · 2 papers
- Meta · 2 papers
- Seoul National University · 2 papers
Use it
Download BibTeX of 35 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Structure-based 3D molecule generation (107 papers)
- Diffusion for molecular conformations (67 papers)
- Graph models for retrosynthesis (10 papers)
- Structure-based molecule design with diffusion (48 papers)
Elsewhere in the topic
- Generative protein structure design (142 papers)
- Prediction-powered and conformal inference (129 papers)
- Statistical analysis of ML practice (122 papers)
- Neuro-symbolic probabilistic layers (120 papers)
- Sparse graphical and tensor models (97 papers)
- Flow matching generative modeling (97 papers)
- Machine learning for physical simulation (96 papers)
- Multimodal trajectory prediction (95 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).