AI-MI researchers continued to make their mark at the International Conference on Machine Learning (ICML 2026) in Seoul. Kilian Weinberger, serving as ICML 2016 program co-chair, led the selection process for this year’s Test of Time Award; his Cornell group also presented a main-track poster on diffusion-model retrieval and a workshop paper on reused-expert mixtures. Tess Smidt’s Atomic Architects group showed two main-track posters — on phonon fine-tuning for machine-learned interatomic potentials, and on fast Clebsch-Gordan tensor products — plus two papers at the Generative and Agentic AI for Biology workshop. Jennifer Sun presented a main-track poster on evaluating agentic optimization over large codebases (with Yisong Yue) and a workshop paper on open-ended scientific discovery.
Great to see AI-MI researchers shaping the conversation at the intersection of AI and physical science.
Source: ICML 2026 Virtual Program