Research map / Probabilistic and scientific modeling / Molecules and proteins
Equivariant Flow Matching with Hybrid Probability Transport for 3D Molecule Generation
Yuxuan Song, Jingjing Gong, Minkai Xu, Ziyao Cao, Yanyan Lan, Stefano Ermon, Hao Zhou, Wei-Ying Ma
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 equivariant OT (EOT) coordinate path instead of vanilla OT (ablation) | EquiFM · QM9 | molecule stability (%): 84.9 -> 88.3 | Improved |
| use EOT instead of VP path on coordinates (ablation) | EquiFM · QM9 | molecule stability (%): 81.6 -> 88.3 | Improved |
| use vanilla OT instead of VP path on coordinates (ablation) | EquiFM · QM9 | molecule stability (%): 81.6 -> 84.9 | Improved |
| use VP cosine instead of VP linear path on atom types (ablation) | EquiFM · QM9 | molecule stability (%): 88.3 -> 84.7 | Hurt |
| use VP polynomial instead of VP linear path on atom types (ablation) | EquiFM · QM9 | molecule stability (%): 88.3 -> 83.4 | Hurt |
| use OT instead of VP linear path on atom types (ablation) | EquiFM · QM9 | molecule stability (%): 88.3 -> 77.1 | Hurt |
| use vanilla OT instead of EOT on coordinates (ablation) | EquiFM · QM9 | atom stability (%): 98.9 -> 98.7 | Hurt |
| use EquiFM instead of EDM-Bridge | EDM · QM9 | molecule stability (%): 84.6 -> 88.3 | Improved |
| use EquiFM flow matching with Dopri5 instead of EDM diffusion | EDM · QM9 | molecule stability: 0.820 -> 0.883 | Improved |
| use EquiFM instead of EDM for conditional generation | EDM · QM9 polarizability alpha | MAE: 2.76 -> 2.41 | Improved |
| use EquiFM instead of EDM for conditional generation | EDM · QM9 dipole moment mu | MAE: 1.111 -> 1.106 | No clear effect |
| use EquiFM instead of EDM for conditional generation | EDM · QM9 heat capacity Cv | MAE: 1.101 -> 1.033 | Improved |
Closest papers by meaning
- Accelerating 3D Molecule Generation via Jointly Geometric Optimal Transport (ICLR 2025 · similarity 0.91)
- Equivariant Neural Diffusion for Molecule Generation (NeurIPS 2024 · similarity 0.91)
- Unified Generative Modeling of 3D Molecules with Bayesian Flow Networks (ICLR 2024 · similarity 0.90)
- Scalable Non-Equivariant 3D Molecule Generation via Rotational Alignment (ICML 2025 · similarity 0.90)
- Controlled Generation with Equivariant Variational Flow Matching (ICML 2025 · similarity 0.89)
- Navigating the Design Space of Equivariant Diffusion-Based Generative Models for De Novo 3D Molecule Generation (ICLR 2024 · similarity 0.89)
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.
