Research map / Architectures and efficiency / RNNs and normalization
May the Force be with You: Unified Force-Centric Pre-Training for 3D Molecular Conformations
Rui Feng, Qi Zhu, Huan Tran, Binghong Chen, Aubrey Toland, Rampi Ramprasad, Chao Zhang
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 zero-force and de-noising regularizations (ablation) | ET-ORE · MD17 Aspirin | force MAE: 4.2 -> 1.0 | Improved |
| add zero-force and de-noising regularizations (ablation) | ET-ORE · MD17 Aspirin | stability (ps): 94 -> 300 | Improved |
| add zero-force and de-noising regularizations (ablation) | ET-ORE · MD17 Naphthalene | stability (ps): 25 -> 300 | Improved |
| add zero-force and de-noising regularizations (ablation) | ET-ORE · MD17 Ethanol | h(r) MAE: 0.10 -> 0.03 | Improved |
| add zero-force and de-noising regularizations (ablation) | ET-ORE · MD17 Ethanol | stability (ps): 300 -> 300 | No clear effect |
| pre-train with forces only (ET-ORE) instead of training from scratch (ablation) | TorchMDNet · MD17 Aspirin | force MAE: 7.4 -> 4.2 | Improved |
| pre-train with forces only (ET-ORE) instead of training from scratch (ablation) | TorchMDNet · MD17 Aspirin | stability (ps): 102 -> 94 | No clear effect |
| full force-centric pre-training before fine-tuning (ablation) | TorchMDNet · MD17 Aspirin | force MAE: 7.4 -> 1.0 | Improved |
| full force-centric pre-training before fine-tuning (ablation) | TorchMDNet · MD17 Naphthalene | stability (ps): 50 -> 300 | Improved |
| pre-train with force-centric objective on diverse conformations (ablation) | TorchMDNet · QM9 HOMO | MAE (meV): 20.3 -> 16.8 | Improved |
| pre-train with force-centric objective on diverse conformations (ablation) | TorchMDNet · QM9 HOMO-LUMO gap | MAE (meV): 36.1 -> 26.4 | Improved |
| force-supervised off-equilibrium pre-training instead of de-noising pre-training | NoisyNode · QM9 HOMO-LUMO gap | MAE (meV): 24.5 -> 26.4 | Hurt |
Closest papers by meaning
- Learning 3D Anisotropic Noise Distributions Improves Molecular Force Fields (NeurIPS 2025 · similarity 0.89)
- Sliced Denoising: A Physics-Informed Molecular Pre-Training Method (ICLR 2024 · similarity 0.89)
- Beyond Atoms: Enhancing Molecular Pretrained Representations with 3D Space Modeling (ICML 2025 · similarity 0.89)
- 3D Denoisers Are Good 2D Teachers: Molecular Pretraining via Denoising and Cross-Modal Distillation (AAAI 2025 · similarity 0.88)
- The dark side of the forces: assessing non-conservative force models for atomistic machine learning (ICML 2025 · similarity 0.87)
- Swallowing the Bitter Pill: Simplified Scalable Conformer Generation (ICML 2024 · 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.
