Saved in:
Bibliographic Details
Main Authors: Serrano, Louis, Han, Jiequn, Oyallon, Edouard, Ho, Shirley, Morel, Rudy
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.00884
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917239123345408
author Serrano, Louis
Han, Jiequn
Oyallon, Edouard
Ho, Shirley
Morel, Rudy
author_facet Serrano, Louis
Han, Jiequn
Oyallon, Edouard
Ho, Shirley
Morel, Rudy
contents Neural operators have shown promise in learning solution maps of partial differential equations (PDEs), but they often struggle to generalize when test inputs lie outside the training distribution, such as novel initial conditions, unseen PDE coefficients or unseen physics. Prior works address this limitation with large-scale multiple physics pretraining followed by fine-tuning, but this still requires examples from the new dynamics, falling short of true zero-shot generalization. In this work, we propose a method to enhance generalization at test time, i.e., without modifying pretrained weights. Building on DISCO, which provides a dictionary of neural operators trained across different dynamics, we introduce a neural operator splitting strategy that, at test time, searches over compositions of training operators to approximate unseen dynamics. On challenging out-of-distribution tasks including parameter extrapolation and novel combinations of physics phenomena, our approach achieves state-of-the-art zero-shot generalization results, while being able to recover the underlying PDE parameters. These results underscore test-time computation as a key avenue for building flexible, compositional, and generalizable neural operators.
format Preprint
id arxiv_https___arxiv_org_abs_2602_00884
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Test-time Generalization for Physics through Neural Operator Splitting
Serrano, Louis
Han, Jiequn
Oyallon, Edouard
Ho, Shirley
Morel, Rudy
Machine Learning
Neural operators have shown promise in learning solution maps of partial differential equations (PDEs), but they often struggle to generalize when test inputs lie outside the training distribution, such as novel initial conditions, unseen PDE coefficients or unseen physics. Prior works address this limitation with large-scale multiple physics pretraining followed by fine-tuning, but this still requires examples from the new dynamics, falling short of true zero-shot generalization. In this work, we propose a method to enhance generalization at test time, i.e., without modifying pretrained weights. Building on DISCO, which provides a dictionary of neural operators trained across different dynamics, we introduce a neural operator splitting strategy that, at test time, searches over compositions of training operators to approximate unseen dynamics. On challenging out-of-distribution tasks including parameter extrapolation and novel combinations of physics phenomena, our approach achieves state-of-the-art zero-shot generalization results, while being able to recover the underlying PDE parameters. These results underscore test-time computation as a key avenue for building flexible, compositional, and generalizable neural operators.
title Test-time Generalization for Physics through Neural Operator Splitting
topic Machine Learning
url https://arxiv.org/abs/2602.00884