Reproduce “UPS: Efficiently Building Foundation Models for PDE Solving via Cross-Modal Adaptation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Junhong Shen, Tanya Marwah, Ameet Talwalkar

UPS maps physical states to next states across selected one- and two-dimensional PDEs. It combines a Fourier neural operator embedder, text-form PDE metadata, a pretrained language model, and a shared prediction head, trained in two stages. On one tested Burgers viscosity change, UPS-Base reached 0.0022 normalized error with 100 trajectories, outperforming FNO and ORCA; this does not establish general few-shot transfer.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.