Research map / Probabilistic and scientific modeling / Molecules and proteins
Flow for Future: Geometric SE(3)-Equivariant Flow Matching for 3D Trajectory Prediction
Junwei Wu, Yihang Liu, Ruixuan Yu, Jian Sun
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 historical context to the flow state encoder (ablation) | GSE-Flow · MD17 Benzene | ADE: 2.35 -> 1.39 | Improved |
| add position time modulation (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.93 -> 1.39 | Improved |
| add velocity time modulation (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.43 -> 1.39 | Improved |
| add node feature time modulation (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.43 -> 1.39 | Improved |
| use geometry-feature tensorization instead of geometric-only attention (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.67 -> 1.39 | Improved |
| use geometry-feature tensorization instead of feature-only attention (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.80 -> 1.39 | Improved |
| use geometry-feature tensorization instead of summing attention scores (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.59 -> 1.39 | Improved |
| aggregate relative displacements instead of absolute representations (ablation) | GSE-Flow · MD17 Benzene | ADE: 1.45 -> 1.39 | Improved |
| add residual connection (ablation) | GSE-Flow · MD17 Benzene | ADE: 4.07 -> 1.39 | Improved |
| derive vector fields from predicted trajectories instead of predicting them directly (ablation) | GSE-Flow · MD17 Benzene | ADE: 2.50 -> 1.39 | Improved |
| add historical context to the flow state encoder (ablation) | GSE-Flow · CMU MoCap Walk | ADE: 85.84 -> 25.92 | Improved |
| replace deterministic sequence encoders with GSE-Flow flow matching in EGNN, EGHN, EGNO (ablation) | GSE-Flow · MD17 | ADE | Improved |
Closest papers by meaning
- SE(3)-Equivariant Flow Matching with Gaussian Process Priors for Geometric Trajectory Prediction (ICML 2026 · similarity 0.92)
- Metric Flow Matching for Smooth Interpolations on the Data Manifold (NeurIPS 2024 · similarity 0.88)
- Geometric Trajectory Diffusion Models (NeurIPS 2024 · similarity 0.88)
- Flow Equivariant Recurrent Neural Networks (NeurIPS 2025 · similarity 0.86)
- Optimal Flow Matching: Learning Straight Trajectories in Just One Step (NeurIPS 2024 · similarity 0.86)
- Not-So-Optimal Transport Flows for 3D Point Cloud Generation (ICLR 2025 · 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.
