Arash Vahdat ✈️ ICML2025
@ArashVahdat
Research Director, leading fundamental generative AI research (GenAIR) @nvidia research, views are my own.
Proud to be on the organizing team for the debut edition of #EurIPS2025! 🇪🇺
NeurIPS is pleased to officially endorse EurIPS, an independently-organized meeting taking place in Copenhagen this year, which will offer researchers an opportunity to additionally present their accepted NeurIPS work in Europe, concurrently with NeurIPS. Read more in our blog…
GenMol: A Drug Discovery Generalist with Discrete Diffusion from @NVIDIAHealth team arxiv.org/abs/2501.06158
🚀 GenMol is now open‑sourced: you can now train and finetune on your data! It uses masked diffusion + a fragment library to craft valid SAFE molecules, from de novo design to lead optimization. #GenMol #DrugDiscovery #Biopharma
📢 Excited to announce that GenMol is now open-sourced. GenMol: A Drug Discovery Generalist with Discrete Diffusion Paper: arxiv.org/abs/2501.06158 Code: github.com/NVIDIA-Digital…
🚀 GenMol is now open‑sourced: you can now train and finetune on your data! It uses masked diffusion + a fragment library to craft valid SAFE molecules, from de novo design to lead optimization. #GenMol #DrugDiscovery #Biopharma
Our 500+ page AI4Science paper is finally published: Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems. Foundations and Trends® in Machine Learning, Vol. 18, No. 4, 385–912, 2025 nowpublishers.com/article/Detail…
🌞🌞🌞Organized by: @JiajunHe614, @YuanqiD, @zdhnarsil, @helibenhamu, Francisco Vargas, @jmhernandez233, @ArnaudDoucet1, @AnimaAnandkumar and Yunan Yang. Check out our webpage for more details: spigmworkshopv3.github.io
✨✨we have a stellar group of speakers and rising star panelists in the field. Speakers: @ArashVahdat, @RuiqiGao, @chelseabfinn, @andrewgwils, @vdbergrianne and Eric Vanden-Eijnden Panelists: @ValentinDeBort1, @XiangLisaLi2, @hlws_bot and Carles Domingo-enrich
❓This argument can go both ways. (1) there is a blooming interest in RL and probabilistic inference to finetune foundation models; (2) the increasing amount of data seem to suggest we can solely rely on data and model generalizability to achieve most tasks of interest.
🌞🌞🌞 The third Structured Probabilistic Inference and Generative Modeling (SPIGM) workshop is **back** this year with @NeurIPSConf at San Diego! In the era of foundation models, we focus on a natural question: is probabilistic inference still relevant? #NeurIPS2025
10 minutes to the @ArashVahdat talk on limitations of generative models to the #ICML2025 ML4Wireless Workshop! Join us in the West Meeting Room 205-207!
What a lineup! Go hang out with @ArashVahdat and the rest of our team from @nvidia @NVIDIAHealth!
Hope to see you all tomorrow at the GenAI & Bio workshop!! #ICML2025 Schedule: genbio-workshop.github.io/2025/
Introducing La-Proteina, a generative AI model that creates complete atomistic protein structures together with their amino acid sequences, scalable upto 800 residues. Built by @NVIDIA Research, in collaboration with @UniofOxford and @Mila_Quebec La-Proteina addresses a key…
La-Proteina: Atomistic Protein Generation via Partially Latent Flow Matching A new generative model, La-Proteina, has emerged for de novo protein structure design, directly generating fully atomistic structures coupled with their amino acid sequences. This addresses a…
We'll present GenMol tomorrow morning at #ICML2025. Come check out our poster and chat about discrete diffusion and small molecule generation. See you at 11 am, West Exhibition Hall B2-B3 #W-113
Can discrete diffusion-based generative models outperform GPT-based next-token predictive models in molecular tasks? Learn about GenMol, NVIDIA’s new #DrugDiscovery Swiss Army Knife! paper: arxiv.org/abs/2501.06158 blog: developer.nvidia.com/blog/evaluatin… demo: build.nvidia.com/nvidia/genmol-…
Partially-latent flow matching enables sequence-structure codesign of large proteins and functional motif scaffolding. @tomasgeffner @DidiKieran @ZhonglinJC @Oxer22 @sacdallago @karsten_kreis @ArashVahdat
🧬Introducing La‑Proteina: a partially‑latent flow‑matching model that co‑generates sequence + all‑atom structure for proteins up to 800 aa 🧬 Side‑chains live in latents, backbone explicit → 75 % codesign & SOTA motif scaffolds 🔥
Check this out! New target-specific dataset and model for binding affinity prediction! 📚 Paper: arxiv.org/abs/2507.08966 💻DualBind code: github.com/NVIDIA-Digital… 🤗 ToxBench dataset: huggingface.co/datasets/karll…
In collaboration with @schrodinger, we present DualBind trained on ToxBench 📚 8,770 ERα AB-FEP predicted complexes (RMSE ≈ 1 kcal mol⁻¹). At last, ML has the signal to learn true binding physics - DualBind already hits r = 0.84 without shortcuts. How could this reshape…
Check out La-Proteina!
Presenting La-Proteina! A new model for scalable, all-atom protein design 🧬 Backbone + sequence + side-chains, indexed and unindexed atomistic motif scaffolding, scalable up to 800 residues, and more… A thread 🧵
Hello #ICML2025👋, anyone up for a diffusion circle? We'll just sit down somewhere and talk shop. 🕒Join us at 3PM on Thursday July 17. We'll meet here (see photo, near the west building's west entrance), and venture out from there to find a good spot to sit. Tell your friends!
Hybrid explicit-latent flows are the new foundation models for protein structures. La-Proteina shows that one network can design 800-residue, all-atom structures and sequences, then perform well at motif scaffolding. Read more: nvda.ws/4nOjBSL