Research map / Probabilistic and scientific modeling / Dynamical systems
GenCP: Towards Generative Modeling Paradigm of Coupled physics
Tianrun Gao, Haoren Zheng, Wenhao Deng, Haodong Feng, Tao Zhang, Ruiqi Feng, Qianyi Chen, Tailin Wu
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 |
|---|---|---|---|
| use GenCP operator-splitting coupled inference instead of M2PDE | flow matching with FNO* backbone · Turek-Hron | average relative L2 error: 26.77% lower average error | Improved |
| use GenCP instead of best baseline | CNO backbone · Turek-Hron | average relative L2 error: 12.54% lower average error | Improved |
| use GenCP instead of baselines | FNO* backbone · Double Cylinder | average relative L2 error: 34.44% error reduction | Improved |
| use GenCP instead of baselines | CNO backbone · Double Cylinder | average relative L2 error: 42.85% error reduction | Improved |
| use GenCP instead of M2PDE and Surrogate | FNO* backbone · Nuclear-Thermal coupling | average relative L2 error: 49.9% reduction vs M2PDE, 51.7% vs Surrogate | Improved |
| use GenCP instead of M2PDE and Surrogate | CNO backbone · Nuclear-Thermal coupling | average relative L2 error: 58.2% over M2PDE and 78.8% over Surrogates | Improved |
| use GenCP instead of M2PDE | flow matching · Synthetic complex distribution | Wasserstein-1 distance: 25450.5442 -> 0.4928 | Improved |
| use GenCP instead of M2PDE | flow matching · Synthetic easy distribution | Wasserstein-1 distance: 0.5177 -> 0.4366 | Improved |
| use GenCP instead of M2PDE for inference cost | FNO* backbone · Turek-Hron | inference time: 277.20s -> 19.50s | Improved |
| use GenCP instead of Surrogate for inference cost | CNO backbone · Turek-Hron | inference time: 300.25s -> 16.25s | Improved |
| use GenCP decoupled training instead of Joint Training on coupled data | FNO* backbone · Double Cylinder | relative L2 error, u: 0.0110 -> 0.0522 | Hurt |
| use GenCP instead of Surrogate for SDF field | FNO* backbone · Turek-Hron SDF | relative L2 error: 0.0112 -> 0.0081 | No clear effect |
Closest papers by meaning
- M2PDE: Compositional Generative Multiphysics and Multi-component PDE Simulation (ICML 2025 · similarity 0.85)
- Generating Physical Dynamics under Priors (ICLR 2025 · similarity 0.85)
- Physics-Constrained Flow Matching: Sampling Generative Models with Hard Constraints (NeurIPS 2025 · similarity 0.84)
- PhysicsGen: Can Generative Models Learn from Images to Predict Complex Physical Relations? (CVPR 2025 · similarity 0.84)
- Interpretable Solutions for Multi-Physics PDEs Using T-NNGP (AAAI 2025 · similarity 0.84)
- Parametric model reduction of mean-field and stochastic systems via higher-order action matching (NeurIPS 2024 · similarity 0.83)
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.
