Research map / Architectures and efficiency / RNNs and normalization
A Cartesian-3j Framework for Machine Learning Interatomic Potentials
Zemin Xu, Chenyu Wu, Wenbo Xie, Peijun Hu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| swap spherical tensors for irreducible Cartesian tensors in NequIP/Allegro (cNequIP, cAllegro) (ablation) | NequIP, Allegro · 3BPA | energy/force RMSE | No clear effect |
| Cartesianize NequIP at 64(64)-3-3-5 (ablation) | NequIP · 3BPA | speed (s/epoch): 15.8 -> 46.4 | Hurt |
| Cartesianize NequIP at 64(64)-3-3-5 (ablation) | NequIP · 3BPA | memory (MiB): 3968 -> 12486 | Hurt |
| Cartesianize Allegro at 256(64)-3-3-3 (ablation) | Allegro · 3BPA | memory: 5862 -> OOM | Hurt |
| Cartesianize MACE (cMACE) (ablation) | MACE · 3BPA | energy/force RMSE | No clear effect |
| Cartesianize MACE at 64-3-3-2-3 (ablation) | MACE · 3BPA | memory (MiB): 1610 -> 15890 | Hurt |
| Cartesianize MACE at 64-3-3-2-3 (ablation) | MACE · 3BPA | speed (s/epoch): 6.7 -> 27.6 | Hurt |
| add gated nonlinear interactions in the convolution (ablation) | TACE 2-layer OMat24 model · Matbench Discovery thermal conductivity | kappa SRME: 0.210 -> 0.158 | Improved |
| use nonlinear interactions | Cartesian models · 3BPA | extrapolation | Hurt |
| use only RCTP/RCTC at edge level instead of Cartesian-3j symbols | TACE architecture · MD speed | speed | Improved |
| use larger truncation Lmax=4, ell_max=4 in Cartesian space | TACE · memory | memory | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MACE — MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields (NeurIPS 2022) · better on 0 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Higher-Rank Irreducible Cartesian Tensors for Equivariant Message Passing (NeurIPS 2024 · similarity 0.94)
- FlashTP: Fused, Sparsity-Aware Tensor Product for Machine Learning Interatomic Potentials (ICML 2025 · similarity 0.87)
- Physics-Informed Weakly Supervised Learning For Interatomic Potentials (ICML 2025 · similarity 0.87)
- Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials (NeurIPS 2026 · similarity 0.87)
- ELECTRA: A Cartesian Network for 3D Charge Density Prediction with Floating Orbitals (NeurIPS 2025 · similarity 0.87)
- The Importance of Being Scalable: Improving the Speed and Accuracy of Neural Network Interatomic Potentials Across Chemical Domains (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
