AI-MI was well-represented at the International Conference on Machine Learning (ICML 2026, Seoul, July 11–19). Eun-Ah Kim, AI-MI Director, co-organized the “AI for Physics” workshop and moderated its opening panel, “AI4Physics and the Current AI Revolution.” John Thickstun (Cornell CS, AI-MI senior personnel) presented a poster on “Scaling Beyond Masked Diffusion Language Models,” demonstrating that uniform-state and interpolating discrete diffusion language models can be made ~12% more FLOPs-efficient. Yoav Artzi (Cornell CS, AI-MI senior personnel) served on the ICML 2026 Outstanding Paper selection committee.
Source: ICML 2026