Rudy Morel

AI&Physics | Resarch Fellow at @FlatironCCM | Member of @PolymathicAI | ex PhD student at @ENS_Ulm

[Miles Cranmer @MilesCranmer]

For the past two years we've studied a decades-old problem in fluid dynamics: why do some turbulent systems grow 3x faster in the real world than simulations predict?

With some tabletop fluids experiments and a physics foundation model, we finally have some results!!!

👇

[Flatiron Institute @FlatironInst]

#FlatironCCQ quantum physicists and colleagues used tensor networks to tackle quantum physics problems previously thought to be solvable only by quantum computers: https://t.co/QjFug79DYM #science #physics

[Rudy Morel @rdMorel]

A great opportunity to do a postdoc in New York @polymathicai, embedded in a truly multidisciplinary environment (NYU + Flatiron).

Generalization and transfer in scientific foundation models remain underexplored, but will be central to many applications in the coming years.

[PolymathicAI @PolymathicAI]

Interested in building the next generation of foundation models for science?
Join Polymathic in New York (NYU + Flatiron) for a postdoc.
Topics include generalization/transfer, and solar physics.

Gen-transfer: https://t.co/Pup1MoJtjP
Solar-physics: https://t.co/yAUh3Obta2

[Siavash Golkar @SiavashGolkar]

1/12 Introducing MIMIC: a SOTA foundation model trained natively across DNA, RNA and proteins. MIMIC is multimodal and generative: it can use structure, regulation, evolution, and experimental context to infer missing biology or design new sequences.

[Shirley Ho @cosmo_shirley]

If you are at #ICLR at Rio, come see some new @PolymathicAI work led by our awesome teammates!

[Alberto Bietti @albertobietti]

Attention and MLP matrices can store as much knowledge as # params, but in practice, finite samples/compute and long context noise can further limit this capacity. We characterize these trade-offs in a needle-in-a-haystack model.

#ICLR2026 poster Thurs, led by @nurimertvural45

[Matthieu Blanke @MatB_29]

I'm excited to present CASAL, our training-free constrained sampling algorithm, at #ICLR2026! 🌴

Leveraging Wasserstein optimization, CASAL enforces hard constraints on pre-trained diffusion models with minimal bias.

Saturday 10:30 AM – 1:00 PM, P3-#409
https://t.co/oq1Tc16QzK

[Shirley Ho @cosmo_shirley]

Join @PolymathicAI and @EmtiyazKhan and me to explore the world of frontier AI for science using new Bayesian techniques!

We are looking for both postdoctoral fellows and research scientists! 🚀

[Luhuan Wu @hlws_bot]

Let’s share your thoughts on generative ai and probabilistic inference at the SPIGM workshop in Seoul, 🇰🇷!

[Flatiron Institute @FlatironInst]

Are you a grad student or postdoc interested in CMB data analysis and simulations? Apply for our #FlatironCCA CMB Summer School! Applications due April 4: https://t.co/3rwjWV7sw5 #science #astrophysics

[Tanya Marwah @__tm__157]

Neural PDE solvers are often framed as a modeling problem.

But in practice, data generation dominates the cost—we rarely ask, how to allocate the compute for generating the data?

Should we spend it all on the hardest simulations?

[Felix Koehler @felix_m_koehler]

Check out my latest video on simulating the KS Equation in 2D & 3D with JAX: https://t.co/1zCYy6s13T

Including some visually stunning 3D volume render animations created with vape4d.

[Luhuan Wu @hlws_bot]

We introduce a new method, EmbedOpt, for robustly steering protein sequence-to-structure diffusion models to fit experimental data (Cryo-EM, NMR) without training. 🧬📉 @mhli41 @JiequnH @PilarCossio2

EmbedOpt tackles the brittleness of the previous coordinate-space steering methods by optimizing the conditional embedding instead. These embeddings capture rich co-evolutionary signals in protein diffusion models—unlocking a new, robust and semantically meaningful diffusion steering axis.

🚀 Result: Better fitting, wider hyperparam stability, and efficiency enabled by fewer diffusion steps

📄 Preprint: https://t.co/Ir7QcXyIW9

[Louis Serrano @LouisSerrano31]

1/7 Excited to share what I started working on last summer at @PolymathicAI with @rdMorel, before joining Emmi AI!
We propose the first framework to boost test-time generalization for OOD spatiotemporal dynamics. No fine-tuning, just search and operator composition!

[Shirley Ho @cosmo_shirley]

Getting ready to speak at the @worldmodel_26 at Mila alongside @ylecun, @Yoshua_Bengio and @sherryyangML :)

Excited to see you online or in person! 🤪

[Florentin Guth @FlorentinGuth]

🔬 Sci4DL is looking for reviewers! Last edition's submissions were truly outstanding, and I can't wait to see this year's program!
🏆 We have prizes for top reviewers to encourage and reward high-quality reviews. If interested please fill the form below:

[Rudy Morel @rdMorel]

Want to develop ML methods for Physics?

Check out our summer internships in the Center for Computational Mathematics at the Flatiron Institute @FlatironInst

Location: Manhattan, New York
Apply here: https://t.co/FzcRXDjccP
Deadline: Jan 16, 2026

[Nikhil Ghosh @nikhilghosh101]

Sharing our recent work on understanding the mechanisms underlying the empirical success of hyperparameter transfer using μP! (1/11)

with Denny Wu and @albertobietti

[Alberto Bietti @albertobietti]

Our new paper on understanding hyperparameter transfer: when does it work, when is it useful, and a conjecture for why fast transfer happens in practice, backed by extensive experiments.

Check out the 🧵 below by the great @nikhilghosh101 for more details!

[Nolan Koblischke @astro_nolan]

We built a semantic search engine for millions of galaxy images by having LLMs write the captions.

These images are completely unlabeled, but our method enables astronomers to search for rare phenomena via text. Try our app! 🔭👇