Happy to chat more at #NeurIPS2023 if you're interested. Feel free to reach out.

Congratulations to the winners and all the participants! 🎉🎊 If you're interested in helping organize next year's challenge, make sure to reach out!

CrystalLLM is up on arxiv! In the paper, we show that LLMs can be fine-tuned to produce 3D structures that are stable according to density functional theory (DFT). arxiv.org/abs/2402.04379 w/ Anuroop Sriram, Andrea Madotto, Andrew Gordon Wilson, Zack Ulissi 1/6

🧵(1/7) Excited to share that our work, EquiformerV2, has been accepted to #ICLR2024. EquiformerV2 is the state-of-the-art on large-scale atomistic benchmarks -- OC20, OC22, AdsorbML, and ODAC23. Joint work with Brandon Wood, Abhishek Das from FAIR Chemistry and Prof. Tess Smidt Paper:

Paper shared on arXiv showing catalyst AI/ML models trained on datasets like FAIR Chemistry can generalize to solid solutions like high entropy alloys (HEA)! This is exciting because the design space of HEAs (with >5 components) is combinatorially large. arxiv.org/abs/2403.09811

🧵(1/9) Introducing DeNS, an auxiliary training objective that generalizes denoising to non-equilibrium atomistic structures, improving the performance of equivariant force fields such as Equiformer. Joint work with Prof. Tess Smidt and Abhishek Das from FAIR Chemistry Preprint:

It’s here! Meet Llama 3, our latest generation of models that is setting a new standard for state-of-the art performance and efficiency for openly available LLMs. Key highlights • 8B and 70B parameter openly available pre-trained and fine-tuned models. • Trained on more

⚛️⚛️⚛️

NEW: ALL SEVEN CONTINENTS HAVE STUDENT GAZA SOLIDARITY PROTESTS 🤯

Equivariant models have become a natural choice for modeling atomic systems. Check out our latest video to learn more! youtube.com/watch?v=Y6Jwgq…

HIND’S HALL. Once it’s up on streaming all proceeds to UNRWA.

🌟 Make sure to check out the new webpage of our work with interactive visualization and additional results demonstrating generalization. That pre-trained checkpoint is a great starting point for many atomic prediction tasks! #ICLR2024

🚨 New work at the intersection of Diffusion x Catalysis x Optimization accepted at #ICML2024🎉! Introducing AdsorbDiff: Adsorbate Placement via Conditional Denoising Diffusion ⚛️ Paper - arxiv.org/abs/2405.03962 Code - github.com/AdeeshKolluru/… 1/n

One of the most common questions we receive is whether OC20 models can be used for transition-state calculations. In this new work we demonstrate that indeed, even without training on transition-state data, OC20 models can accelerate these calculations out of the box!

Wrapping up the Open Catalyst Intro Series (for now) - how do we evaluate models and address open problems that still remain youtu.be/DYyTYcwaHUk?si… If you'd like to see a deep dive on a topic that wasn't already covered, we are always open for feedback!

Introducing fairchem - our revamped codebase consolidating our AI modeling efforts in chemistry and materials science. fairchem makes it easy to interface with our data, models, demos, and applications - including an easy to use ASE calculator: github.com/FAIR-Chem/fair…

The ocp repo has now rebranded to fairchem. As part of the upgrade, we have made installation, models, data, and tutorials a lot more accessible to new users. Check it out here: github.com/FAIR-Chem/fair….

Announcing our new model for materials! FlowMM... - Generates stable & novel materials efficiently - Predicts crystal structure accurately - Generalizes Riemannian Flow Matching to point clouds w/ periodic boundaries arxiv.org/abs/2406.04713 Ricky T. Q. Chen Anuroop Sriram Brandon Wood

Data generation through simulation is one reason why the whole idea of ML for science is so exciting.

A few of us will be at the AC conference this week. We will be sharing some of our ongoing computational-experimental efforts for electrocatalysis. Make sure to come and chat with Jehad Abed, Muhammed Shuaibi, and Larry Zitnick.