AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

MSPT: Efficient Large-Scale Physical Modeling via Parallelized Multi-Scale Attention

Pedro M. P. Curvo, Jan-Willem van de Meent, Maksim Zhdanov

CVPR 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer neural operators for PDEs · 3 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 Transolver global slices with MSPT patch attention plus pooled supernodesMSPT · Navier-Stokesrelative L2 (x1e-2): 9.00 -> 6.32Improved
MSPT versus Transolver on ElasticityMSPT · Elasticityrelative L2 (x1e-2): 0.64 -> 0.48Improved
MSPT versus Transolver on DarcyMSPT · Darcyrelative L2 (x1e-2): 0.57 -> 0.63Hurt
MSPT versus Erwin on ElasticityMSPT · Elasticityrelative L2 (x1e-2): 0.34 -> 0.48Hurt
MSPT versus Transolver on ShapeNet-Car volumeMSPT · ShapeNet-Car volumerelative L2 (x1e-2): 2.07 -> 1.89Improved
MSPT versus Transolver on AhmedML volumeMSPT · AhmedML volumerelative L2 (x1e-2): 2.05 -> 2.04No clear effect
MSPT versus Transolver on AhmedML surfaceMSPT · AhmedML surfacerelative L2 (x1e-2): 3.45 -> 3.22Improved
vary number of patches K from 32 to 1024 (ablation)MSPT · ShapeNet-Car test losstest loss (x1e-2): 6.08 (K=32), 6.83 (K=128), 5.99 (K=1024)Mixed
use max pooling instead of mean pooling for supernodes (ablation)MSPT · ShapeNet-Car test losstest lossHurt
use learned linear projection instead of mean pooling (ablation)MSPT · ShapeNet-Car test losstest lossNo clear effect
increase supernodes per patch Q above 1 (ablation)MSPT · ShapeNet-Car test losstest lossImproved
MSPT instead of Transolver for memory at 500k pointsMSPT · 500k points efficiencypeak memory (GB): 42.8 -> 26.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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