Skip to content
AI News HubLIVE
Original source2 min read

Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation

Summary

This study reveals catastrophic forgetting in Walrus-1.3B during a two-stage sim-to-real adaptation: finetuning on simulated cylinder flow for just 20 epochs increases one-step errors up to 35x, and subsequent real-data adaptation further erodes simulation performance. Replaying only 1% simulated data recovers most lost simulation performance while retaining effective adaptation.

SourcearXiv Machine LearningAuthor: Haaris Mian, Miles Cranmer, Alessandro Favero, Payel Mukhopadhyay
Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 29 Oct 2026]

Title:Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation

View a PDF of the paper titled Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation, by Haaris Mian and 3 other authors

View PDF HTML (experimental)

Abstract:Scientific foundation models (SciFMs) have achieved remarkable performance on a wide range of PDEs, including the ability to efficiently adapt to unseen data distributions with additional finetuning. As with all neural architectures, finetuning such models comes with the risk of catastrophic forgetting, whereby adapting to new tasks degrades prior knowledge of the model. We study catastrophic forgetting in Walrus-1.3B, a SciFM for continuum dynamics, over the course of a two-stage sim-to-real adaptation: first finetuning on simulated flow around a cylinder and then adapting to experimental data of the same system. We find that finetuning on the simulations rapidly degrades rollout performance on the original pretraining datasets, with one-step errors increasing up to 35x after just 20 epochs. Subsequent adaptation to real data further erodes performance on the simulated task. We characterize the potential drivers of forgetting, such as dataset dependence and task representation similarity. We then show that replaying as little as 1% simulated data during real-data finetuning recovers most of the lost simulation performance while retaining effective adaptation to the experimental dynamics. Our results identify catastrophic forgetting as a practical obstacle to sim-to-real adaptation of SciFMs and show that lightweight replay strategies significantly improve retention.

Comments: Accepted at the Sim2Science workshop, NeurIPS 2026. Code: this https URL

Subjects:

Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)

Cite as: arXiv:2610.40857 [cs.LG]

(or arXiv:2610.40857v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2610.40857

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Haaris Mian [view email] [v1] Thu, 29 Oct 2026 18:36:42 UTC (17,569 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation, by Haaris Mian and 3 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.LG

new | recent | 2026-10

Change to browse by:

cs physics physics.flu-dyn

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

IArxiv recommender toggle

IArxiv Recommender (What is IArxiv?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Finetuning Walrus-1.3B on simulated cylinder flow rapidly degrades rollout performance on original pretraining datasets, with one-step errors rising up to 35x after only 20 epochs.
  • Subsequent adaptation to experimental data further erodes performance on the simulated task, compounding forgetting across the two-stage process.
  • Replaying as little as 1% simulated data during real-data finetuning recovers most of the lost simulation performance while preserving effective adaptation.
  • The work identifies catastrophic forgetting as a practical obstacle to sim-to-real adaptation of scientific foundation models and validates lightweight replay as a mitigation.

Highlights and analysis are generated automatically and may contain errors. Check the original source.

More recent stories on this topic · Models