Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Physics-Constrained Fine-Tuning of Flow-Matching Models for Generation and Inverse Problems
Jan Tauberschmidt, Sophie Fellenz, Sebastian Josef Vollmer, Andrew B. Duncan
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 |
|---|---|---|---|
| fine-tune with joint solution-parameter flow instead of vanilla adjoint matching (frozen phi) (ablation) | adjoint matching fine-tuning · Stokes lid-driven cavity | MMD_alpha: 0.22-0.28 -> 0.07-0.13 | Improved |
| fine-tune with joint solution-parameter flow instead of Base AM variants (ablation) | adjoint matching fine-tuning · Stokes lid-driven cavity | weak residual | No clear effect |
| fine-tune with joint solution-parameter flow instead of Base AM variants (ablation) | adjoint matching fine-tuning · Helmholtz | MMD_x: 0.15 -> 0.07 | Improved |
| fine-tune with adjoint matching on weak-form residuals | base flow matching model · Helmholtz | weak residual (rel): weak residual down to 4.3x10^0 from 1.5x10^1 | Improved |
| increase lambda_x = lambda_alpha at lambda_f = 0 (ablation) | adjoint matching fine-tuning · Darcy flow | PDE residual and inferred permeability diversity | Mixed |
| increase regularization lambda_f (ablation) | adjoint matching fine-tuning · Darcy flow | MMD_x and PDE residual | Mixed |
| fine-tune with regularization lambda_f=1.0 (ablation) | adjoint matching fine-tuning · Darcy flow noisy pressure | sample-specific fidelity | Improved |
| use PBFM baseline instead of adjoint matching | physics-constrained fine-tuning · Stokes lid-driven cavity | convergence: strong residuals 1.15x10^1 | Hurt |
| fine-tune with PBFM or FM+ECI instead of our method | physics-constrained fine-tuning · Linear elasticity BC misspecification | MMD_x: MMD_x 0.15 (ours) vs 0.92 (PBFM) and 1.16 (FM+ECI) | Hurt |
| use scaled memoryless noise schedule (kappa > 0) | adjoint matching · PDE models in pixel space | fine-tuning stability | Improved |
Closest papers by meaning
- Physics-Constrained Flow Matching: Sampling Generative Models with Hard Constraints (NeurIPS 2025 · similarity 0.93)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.90)
- D-Flow: Differentiating through Flows for Controlled Generation (ICML 2024 · similarity 0.88)
- Physics-Informed Diffusion Models (ICLR 2025 · similarity 0.88)
- Gradient-Free Generation for Hard-Constrained Systems (ICLR 2025 · similarity 0.88)
- Equilibrium Matching: Generative Modeling with Implicit Energy-Based Models (NeurIPS 2026 · 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.
