Does a smaller latent space lead to worse generation in latent diffusion models? Not necessarily! We show that LDMs are extremely robust to a wide range of compression rates (10-1000x) in the context of physics emulation. We got lost in latent space. Join us 👇

The final days of summer are upon us, and it is bittersweet to say goodbye to our great group of @PolymathicAI interns! 😭 @JacopoTeneggi @cskokgibbs @astro_nolan @CristianaD2202 @LouisSerrano31 @rachelczhang Here are a few pics to remind us all the fun we had! (and hold your

Thanks everyone who came to my poster @icmlconf. I'm so happy to feel the excitement about using physics simulations to test and train science agents.

Paper: https://t.co/xhIpdXWJkL Code:

Very excited to present Zebra at ICML today! Zebra is an LLM trained with in-context examples on physics data to directly adapt to new spatiotemporal dynamics, bypassing gradient steps at inference. Come check out my poster at 11:00 AM, West Hall B2-B3 #W-106.

For evolving unknown PDEs, ML models are trained on next-state prediction. But do they actually learn the time dynamics: the "physics"? Check out our poster (W-107) at #ICML2025 this Wed, Jul 16. Our "DISCO" model learns the physics while staying SOTA on next states prediction!

We release AB-UPT, a novel method to scale neural surrogates to CFD meshes beyond 100 million of mesh cells. AB-UPT is extensively tested on the largest publicly available datasets. 📄 https://t.co/xGQxhU8PuJ 🤗 https://t.co/WIirIMyNNd 💻 https://t.co/VuXjboZ0Xo

Given an image, JAFAR builds high-res queries at the target resolution and low-res, semantically enriched keys using spatial feature modulation to power a cross-resolution attention mechanism that interpolates the low-resolution features from the foundation vision encoder (3/n)

🚀Thrilled to introduce JAFAR—a lightweight, flexible, plug-and-play module that upsamples features from any Foundation Vision Encoder to any desired output resolution (1/n) Paper : https://t.co/le4pF8rVXH Project Page: https://t.co/rLW3jbin3O Github: https://t.co/1AL7ElibHf

Today marks a big milestone for us at Emmi AI. We’ve raised a €15M seed round, backed by 3VC, Speedinvest, Serena, and PUSH. Let’s build the future of Physics AI together!

me trying to cut my ICML rebuttal down to <5000 characters

"Learning a Neural Solver for Parametric PDEs to Enhance Physics-Informed Methods" by Lise Le Boudec, @EBezenac, @LouisSerrano31, Ramon Daniel Regueiro-Espino, @yuanyinnn, Patrick Gallinari in collaboration with ETH Zürich, @CriteoAILab, @valeoai, INRIA ➡️ https://t.co/H5zRcd0enw

We present Hypernetwork Fields. We estimate the entire convergence trajectory for hypernetworks by introducing an extra variable representing the state of convergence. We show results for our model estimating DreamBooth parameters. 1/N🧵

Proud to announce ArchesWeatherGen, an #opensource flow matching model for weather forecasting. ArchesWeatherGen surpasses IFS ENS (the reference operational model run by ECMWF) and NeuralGCM (hybrid physics-ML). Paper: https://t.co/f9tkI79Kia Code: https://t.co/9O8aXnBczT

Here's one to read on your flight to #NeurIPS2024! A flow-matching transformer model in function space! This model has all the advantages of neural fields: resolution-free generation and domain-agnostic architecture, while obtaining strong results on ImageNet-256 and Objaverse!

@MustafaShukor1 and @LouisSerrano31 presented their poster during day 2:

♟️Mastering Board Games by External and Internal Planning with Language Models♟️ I'm happy to finally share https://t.co/1Hc2p883Wf TL;DR: In chess, our planning agents effectively reach grandmaster-level strength with a comparable search budget to that of human players!

We release AIMv2, a major step in scaling vision encoders. Properly scaling vision encoders has been challenging and lagging, compared to LLMs. The main bottleneck is training and evaluating on single image modality, (1/n)

Super hyped to share NeuralDEM -- the first real-time simulation of industrial particulate flows. NeuralDEM replaces Discrete Element Method (DEM) routines and coupled (CFD-DEM) multiphysics simulations. 🧵 📜: https://t.co/JH4PDpth5g 🖥️: https://t.co/VEsawzd9IV

"GEPS: Boosting Generalization in Parametric PDE Neural Solvers through Adaptive Conditioning" by @ArmandKassai, @JorgeMifsut, @yuanyinnn, Jean Noël Vittaut, Patrick Gallinari accepted as conference paper at #NeurIPS2024 ! ➡️ https://t.co/ztSKO05ogo 🖥️ https://t.co/Pz4YSrEYLw