🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

🚨 Attention aspiring PhD students: Meta / FAIR is looking for candidates for a joint academic/industry PhD! 🚨 Among others, the CodeGen team is looking for candidates to work on world models for code, discrete search & continuous optimization methods for long-term planning,

Interested in working with a highly collaborative, interdisciplinary team to push the state of the art of generative AI for materials design? Join us as an intern by applying through this link! We are the team behind the MatterGen and MatterSim models from Microsoft Research AI

Introducing Meta’s Open Materials 2024 (OMat24) Dataset and Models! All under permissive open licenses for commercial and non-commercial use! Paper: arxiv.org/abs/2410.12771 Dataset: huggingface.co/datasets/fairc… Models: huggingface.co/fairchem/OMAT24 🧵1/x

📢I’ll be admitting multiple PhD students this winter to Columbia University 🏙️ in the most exciting city in the world! If you are interested in dissecting modern deep learning systems to probe how they work, advancing AI safety, or automating data science, apply to my group.

My ICLR talk “How Do We Build a General Intelligence?” is now online! youtube.com/watch?v=HEp4TO…

LLMs are highly constrained biological sequence optimizers. In new work led by Angelica Chen & Samuel Stanton , we show how to drive an active learning loop for protein design with an LLM. 1/

I’m excited to share our latest work on generative models for materials called FlowLLM. FlowLLM combines Large Language Models and Riemannian Flow Matching in a simple, yet surprisingly effective way for generating materials. arxiv.org/abs/2410.23405 Benjamin Kurt Miller Ricky T. Q. Chen Brandon Wood

I am recruiting Ph.D. students for my new lab at New York University! Please apply, if you want to work with me on reasoning, reinforcement learning, understanding generalization and AI for science. Details on my website: izmailovpavel.github.io. Please spread the word!

I’ll be at NeurIPS this week! If you want to chat about generative modeling for scientific data, LLM uncertainty quantification, or anything else I’ve worked on, please reach out!

New model trained on new dataset of nearly a million evolving antibody families at AIDrugX workshop Sunday at 4:20 pm (#76) #Neurips! Collab between Andrew Gordon Wilson and BigHat Biosciences. Stay tuned for full thread on how we used the model to optimize antibodies in the lab in coming days!

Diffusion models produce high quality outputs and have powerful guidance mechanisms. Recently, discrete diffusion has shown strong language modeling. ❓Can we adapt guidance mechanisms for discrete diffusion models to enable more controllable discrete generation? 1/13 🧵

This project has something for everyone to love! Multi-sequence language models, twisted SMC, wet lab experiments. Go check it out.

Still more amazing professors signing on as potential coadvisors for our FutureHouse Fellowship. Kind of insane: -Debbie Marks (Harvard) Debora Marks -Brian Hie (Stanford/Arc) Brian Hie -Jonathan Gootenberg and Omar Abudayyeh (Jonathan Gootenberg Omar Abudayyeh) -Anima Anandkumar

To trust LLMs in deployment (e.g., agentic frameworks or for generating synthetic data), we should predict how well they will perform. Our paper shows that we can do this by simply asking black-box models multiple follow-up questions! w/ Marc Finzi and Zico Kolter 1/ 🧵

Got a diffusion model? What if there were a way to: - Get SOTA text-to-image prompt fidelity, with no extra training! - Steer continuous and discrete (e.g. text) diffusions - Beat larger models using less compute - Outperform fine-tuning - And keep your stats friends happy !?

Next Tues (3/18) at 4PM ET we will have Alan Amin present "Bayesian Optimization of Antibodies Informed by a Generative Model of Evolving Sequences" Paper: arxiv.org/abs/2412.07763 Sign up on our website for zoom links!

Happening today at 4PM ET!

This paper is for all the discrete diffusion enthusiasts out there! We shed some light on why simple masking diffusion usually outperforms other noise processes.

Autoregressive models are too restrictive by forcing a fixed generation order, while masked diffusion is wasteful as it fits all possible orders. Can our model dynamically decide the next position to generate based on context? Learn more in our ICML paper arxiv.org/abs/2503.05979

Really interesting article by Leash Bio - I think we (I?) should hold our horses thinking drug discovery will be solved by structure prediction as a foundation model Turns out a simple architecture + really scaling binding data = v. v. strong binding affinity predictor