AI Paper Atlas

Research map / Probabilistic and scientific modeling / Differential equations and PDEs

Structure-Preserving Learning Improves Geometry Generalization in Neural PDEs

Benjamin D Shaffer, Shawn Koohy, Brooks Kinch, M. Ani Hsieh, Nathaniel Trask

ICML 2026 · Probabilistic and scientific modeling › Differential equations and PDEs · approach: Neural PDE solvers · 0 citations (OpenAlex)

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
replace direct regression with Geo-NeW structure-preserving implicit FEM model (ablation)inducing-point transformer encoder · NS2d-c++ ID, point-cloud inputprediction error: 8.82e-2 -> 6.76e-2Improved
use mesh-based input encoding instead of point-cloud (ablation)Geo-NeW · NS2d-c++ IDprediction error: 6.76e-2 -> 5.17e-2Improved
use mesh-based input encoding instead of point-cloud (ablation)baseline inducing-point encoder · NS2d-c++ IDprediction error: 8.82e-2 -> 7.21e-2Improved
use Geo-NeW instead of Transolvergeometry extrapolation · Poly-Poisson OODnormalized L2 error (x1e-2): 7.04 -> 2.14Improved
use Geo-NeW instead of Transolvergeometry extrapolation · NS2d-c++ OODnormalized L2 error (x1e-2): 91.4 -> 42.2Improved
use Geo-NeW instead of strongest baselinemild extrapolation · NS2d-c++ OOD (mild)normalized L2 error (x1e-2): 13.1 -> 7.87Improved
use Geo-NeW instead of Transolverin-distribution · NS2d-c++ IDnormalized L2 error (x1e-2): 4.04 -> 1.93Improved
use Geo-NeW instead of Transolverin-distribution · NS2d-cnormalized L2 error (x1e-2): 0.63 -> 1.10Hurt
use Geo-NeW instead of LaMOstandard benchmark · NACA (derived quantity)normalized L2 error (x1e-2): 0.41 -> 0.82Hurt
use Geo-NeW instead of LaMOstandard benchmark · Pipenormalized L2 error (x1e-2): 0.38 -> 0.11Improved
use Geo-NeW instead of LaMOstandard benchmark · Elasticity (derived quantity)normalized L2 error (x1e-2): 0.50 -> 0.35Improved
enforce Dirichlet conditions exactly via boundary partitions (ablation)Geo-NeW · NS2d-c boundary errornormalized L2 boundary error (u): 2.22e-3 -> 0.00Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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