Test-time Generalization for Physics through Neural Operator Splitting

Louis Serrano, Jiequn Han, Edouard Oyallon, Shirley Ho, Rudy Morel
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:109205-109235, 2026.

Abstract

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 multi 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.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-serrano26a, title = {Test-time Generalization for Physics through Neural Operator Splitting}, author = {Serrano, Louis and Han, Jiequn and Oyallon, Edouard and Ho, Shirley and Morel, Rudy}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {109205--109235}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/serrano26a/serrano26a.pdf}, url = {https://proceedings.mlr.press/v306/serrano26a.html}, abstract = {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 multi 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.} }
Endnote
%0 Conference Paper %T Test-time Generalization for Physics through Neural Operator Splitting %A Louis Serrano %A Jiequn Han %A Edouard Oyallon %A Shirley Ho %A Rudy Morel %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-serrano26a %I PMLR %P 109205--109235 %U https://proceedings.mlr.press/v306/serrano26a.html %V 306 %X 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 multi 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.
APA
Serrano, L., Han, J., Oyallon, E., Ho, S. & Morel, R.. (2026). Test-time Generalization for Physics through Neural Operator Splitting. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:109205-109235 Available from https://proceedings.mlr.press/v306/serrano26a.html.

Related Material