We’re bringing natural-language control to any scientific generator. We are very excited to introduce LLM-to-Flow Distillation (LFD), a method to adapt scientific generators to high-level natural-language discovery goals. Starting from textual goals like “design peptides with promising anticancer activity” or “design small molecules binding to the D2 dopamine receptor”, LFD distills high-level goal-related knowledge from LLMs into scientific generators through iterative DPO on LLM-generated sub-goals. 🔥 LFD does not need task-specific proxy rewards, works with any frontier LLM via API-access, and outperforms state-of-the-art reward-guided methods such as FlowGRPO and DiffusionNFT on held-out computational evaluations. (1/6)

Generative models can’t discover what they can’t reach. We’re excited to introduce ActFlow: a continued pre-training scheme that actively expands the valid design space reachable by flow and diffusion models. We call this generable set expansion — a new learning principle for out-of-distribution generative modeling, and a step toward evolvable search spaces for scientific discovery. (1/5)

I’m at #ICML2026 presenting several works on flow and diffusion models for discovery: from theory and algorithms to molecules, peptides, and proteins. These works have been done with amazing students and collaborators at @ETH_AI_Center, Caltech, KTH, and UPenn, including @maltefranke_ , @KimonProto6689, @yapinghsieh, @cristianpjensen, Sven Gutjahr, Luca Schaufelberger, Zifan Wang, @_sophia_tang_, Bruce Lee, Chenghao Liu, @kjelljorner, Karl Henrik Johansson, @pranamanam, @yisongyue, and @arkrause. Feel free to reach out if you’d like to chat these days at ICML!