Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
BLIPs: Bayesian Learned Interatomic Potentials
Dario Coscia, Pim De Haan, Max Welling
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 BLIP variational adaptive dropout | GNN · N-body | MSE x1e-1: 0.121 -> 0.092 | Improved |
| add BLIP variational adaptive dropout | EGNN · N-body | MSE x1e-1: 0.069 -> 0.060 | Hurt |
| use BLIP instead of MC Dropout (p=0.5) | EGNN · N-body | NLL: 0.91 -> -12.52 | Improved |
| use Deep Ensemble instead of BLIP | EGNN · N-body | NLL: -12.52 -> -16.22 | Improved |
| use BLIP instead of Deep Ensemble | PaiNN · Ammonia OOD | forces MAE: 48.03 -> 41.93 | Improved |
| use BLIP instead of Deep Ensemble | PaiNN · Ammonia OOD | energy MAE: 11.71 -> 11.51 | Improved |
| fine-tune with BLIP and random structure selection | ORB v3 · MATPES PBE | energy MAE: 111.83 -> 96.65 | Improved |
| fine-tune with BLIP and random structure selection | ORB v3 · MATPES PBE | forces MAE: 110.84 -> 105.12 | Improved |
| select active-learning structures by BLIP uncertainty instead of randomly (ablation) | ORB v3 + BLIP · MATPES PBE | energy MAE: 96.65 -> 91.13 | Improved |
| select active-learning structures by BLIP uncertainty instead of randomly (ablation) | ORB v3 + BLIP · MATPES PBE | forces MAE: 105.12 -> 103.85 | Improved |
| use BLIP instead of Deep Ensemble for calibration | ORB v3 · SiO2 (1024 samples) | ECE: 0.247 -> 0.032 | Improved |
| use BLIP instead of Deep Ensemble for variance-error rank correlation | ORB v3 · SiO2 (1024 samples) | Spearman correlation: 0.676 -> 0.602 | Hurt |
Closest papers by meaning
- MLIP Arena: Advancing Fairness and Transparency in Machine Learning Interatomic Potentials via an Open, Accessible Benchmark Platform (NeurIPS 2025 · similarity 0.89)
- Physics-Informed Weakly Supervised Learning For Interatomic Potentials (ICML 2025 · similarity 0.88)
- Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials (NeurIPS 2026 · similarity 0.88)
- Colour me shocked: Exact Molecular Hessians from MLIPs in O(N) time using sparse differentiation! (NeurIPS 2026 · similarity 0.87)
- Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction (ICML 2025 · similarity 0.87)
- Stein Kernelized Molecular Dynamics for Active Learning of Interatomic Potentials (NeurIPS 2026 · 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.
