Physics-informed fine-tuning of foundation models for partial differential equations
Vlad Medvedev, Leon Armbruster, Christopher Straub, Georg Kruse, & Andreas Rosskopf · 2026
Read Physics-informed fine-tuning of foundation models for partial differential equations by Vlad Medvedev, Leon Armbruster, Christopher Straub, Georg Kruse, & Andreas Rosskopf (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15431.txt.
Description
Medvedev and co-authors fine-tune PDE foundation models with physics in the loss, not just data. Pretrained surrogates struggle to adapt when task data is scarce or shifted. Their method adds PDE residuals and boundary rules straight into the fine-tuning goal. On an unseen class of equations, the approach matched data-driven tuning without any solved examples for training. A hybrid mix of physics and data generalized best to out-of-distribution cases with tiny training sets. The results frame a data-efficient path for scientific machine learning.
Catalog record
| CREATOR | Vlad Medvedev, Leon Armbruster, Christopher Straub, Georg Kruse, & Andreas Rosskopf |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36765 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.15431