AI Paper Atlas

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

ICLR 2026 · Probabilistic and scientific modeling › Differential equations and PDEs · approach: Physics-informed operator learning

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 changedWhereMeasuredEffect
fine-tune with joint solution-parameter flow instead of vanilla adjoint matching (frozen phi) (ablation)adjoint matching fine-tuning · Stokes lid-driven cavityMMD_alpha: 0.22-0.28 -> 0.07-0.13Improved
fine-tune with joint solution-parameter flow instead of Base AM variants (ablation)adjoint matching fine-tuning · Stokes lid-driven cavityweak residualNo clear effect
fine-tune with joint solution-parameter flow instead of Base AM variants (ablation)adjoint matching fine-tuning · HelmholtzMMD_x: 0.15 -> 0.07Improved
fine-tune with adjoint matching on weak-form residualsbase flow matching model · Helmholtzweak residual (rel): weak residual down to 4.3x10^0 from 1.5x10^1Improved
increase lambda_x = lambda_alpha at lambda_f = 0 (ablation)adjoint matching fine-tuning · Darcy flowPDE residual and inferred permeability diversityMixed
increase regularization lambda_f (ablation)adjoint matching fine-tuning · Darcy flowMMD_x and PDE residualMixed
fine-tune with regularization lambda_f=1.0 (ablation)adjoint matching fine-tuning · Darcy flow noisy pressuresample-specific fidelityImproved
use PBFM baseline instead of adjoint matchingphysics-constrained fine-tuning · Stokes lid-driven cavityconvergence: strong residuals 1.15x10^1Hurt
fine-tune with PBFM or FM+ECI instead of our methodphysics-constrained fine-tuning · Linear elasticity BC misspecificationMMD_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 spacefine-tuning stabilityImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy