Research map / Graph learning / Graph self-supervised learning
Condition-Aware Graph Flow Matching for Modeling the Distributions of Complex Fluid Systems
Xiaochao Deng, Jie Chen, Xiaogang Deng
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 hierarchical multilevel message passing (HieraGraphNet) (ablation) | CGFM · Wing (3D) | graph-wise Wasserstein-2: 3.5026 -> 1.9265 | Improved |
| add hierarchical message passing (ablation) | CGFM · Wing (3D) | R2: 0.9712 -> 0.9896 | Improved |
| use topology- and geometry-aware coarsening instead of Guillard coarsening (ablation) | CGFM · Wing (3D) | node-wise Wasserstein-2: 0.0141 -> 0.0099 | Improved |
| use TGC instead of Guillard coarsening (ablation) | CGFM · Wing (3D) | graph-wise Wasserstein-2: 2.3697 -> 1.9265 | Improved |
| use TGC instead of Guillard coarsening on TandemFoilSet (ablation) | CGFM · TandemFoilSet CruiseAOA=5 | distribution/sample accuracy | Improved |
| increase number of hierarchy levels (ablation) | CGFM · Wing (3D) | R2, W2 distances | Improved |
| reduce number of inference steps (ablation) | CGFM · Wing (3D) | node-wise Wasserstein-2 | No clear effect |
| reduce number of states per training trajectory (ablation) | CGFM · Ellipse-InDist | distributional error | No clear effect |
| use flow matching with hierarchical graph instead of DDPM-based DGN | CGFM vs DGN · Wing (3D) | inference time (s/sample): 0.1377 -> 0.0981 | Improved |
| sample from learned distribution on GPU instead of numerical solver | CGFM · Wing (3D), 3,000 samples | speed-up over solver: 16.05x (CPU), 606.11x (GPU) | Improved |
| use deterministic model (MuGS-GNN) instead of generative model | MuGS-GNN · Ellipse-InDist | graph-wise Wasserstein-2: 0.9727 vs 0.2734 (CGFM) | Hurt |
Closest papers by meaning
- Learning Distributions of Complex Fluid Simulations with Diffusion Graph Networks (ICLR 2025 · similarity 0.90)
- Elucidating Flow Matching ODE Dynamics via Data Geometry and Denoisers (ICML 2025 · similarity 0.86)
- Diff2Flow: Training Flow Matching Models via Diffusion Model Alignment (CVPR 2025 · similarity 0.86)
- Pairwise Optimal Transports for Training All-to-All Flow-Based Condition Transfer Model (NeurIPS 2025 · similarity 0.85)
- Wavelet Flow Matching for Multi-Scale Physics Emulation (NeurIPS 2026 · similarity 0.85)
- Shortcutting Pre-trained Flow Matching Diffusion Models is Almost Free Lunch (NeurIPS 2025 · similarity 0.85)
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.
