Which architectures are actually winning — from CGCNN and MEGNet to EquiformerV2 and hybrid transformer-graph models — on the Matbench leaderboard Predicting a material property — band gap, formation energy,…
Foundation models, self-driving labs, and LLM research agents — the full stack that compresses 10–20-year timelines into 1–2 years The pace of materials innovation is on a step change. DeepMind’s…