Research map / Probabilistic and scientific modeling / Molecules and proteins
From Evaluation to Design: Using Potential Energy Surface Smoothness Metrics to Guide ML Interatomic Potential Architectures
Ryan Liu, Eric Qu, Tobias Kreiman, Samuel M Blau, Aditi S. Krishnapriyan
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 |
|---|---|---|---|
| add controllable Gaussian smearing width (scale 5) (ablation) | MinDScAIP · BSCT | FSD (full): 97.4 -> 83.1 | Improved |
| add temperature-controlled attention (tau 10) (ablation) | MinDScAIP · BSCT | FSD (full): 97.4 -> 75.5 | Improved |
| add weight decay (ablation) | MinDScAIP · BSCT | FSD (full): 97.4 -> 76.3 | Improved |
| combine smearing and temperature (ablation) | MinDScAIP · BSCT | FSD (full): 97.4 -> 43.2 | Improved |
| combine smearing and temperature (ablation) | MinDScAIP · SPICE test | Energy MAE (meV/atom): 0.09 -> 0.12 | Hurt |
| replace standard kNN with Diff-kNN (ablation) | MinDScAIP (gradient forces) · NVE simulation | Energy drift (meV/atom): 19.10 -> 0.678 | Improved |
| use gradient-based forces instead of direct force head (ablation) | MinDScAIP · SPICE/BSCT | Energy drift / FSD: 2.6e5 -> 0.678; FSD 71.8 -> 97.4 | Mixed |
| scale model from small (3.8M) to large (60M) (ablation) | MinDScAIP · SPICE/BSCT | FSD (full): 80.2 -> 97.4 | Hurt |
| scale model from small (3.8M) to large (60M) (ablation) | MinDScAIP · SPICE test | Forces MAE (meV/A): 6.63 -> 3.02 | Improved |
| strengthen smoothness designs from weak to strong (ablation) | MinDScAIP-30M · Matbench Discovery | kappa_SRME: 0.77 -> 0.49 | Improved |
| lower FSD via smoothness designs (Vanilla to Smear. & Temp.) (ablation) | MinDScAIP · MD22 at 2000K | Max kinetic temperature jump: 9734 -> 490 | Improved |
Closest papers by meaning
- Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction (ICML 2025 · similarity 0.92)
- MLIP Arena: Advancing Fairness and Transparency in Machine Learning Interatomic Potentials via an Open, Accessible Benchmark Platform (NeurIPS 2025 · similarity 0.90)
- Physics-Informed Weakly Supervised Learning For Interatomic Potentials (ICML 2025 · similarity 0.89)
- Higher-Rank Irreducible Cartesian Tensors for Equivariant Message Passing (NeurIPS 2024 · similarity 0.88)
- The dark side of the forces: assessing non-conservative force models for atomistic machine learning (ICML 2025 · similarity 0.87)
- EDBench: Large-Scale Electron Density Data for Molecular Modeling (NeurIPS 2025 · similarity 0.87)
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.
