Binbin Xiong is a Research Scientist in the GenAI unit at Google DeepMind, where he leads and co-leads multiple initiatives in Gemini pretraining data research. His work spans key areas such as data valuation, mixture design, and large-scale pretraining data scaling laws. Previously, Binbin spent several years specializing in applied machine learning and deep learning, with a focus on data-centric modeling and interpretability techniques to enhance Google’s products, including Search, YouTube, Waymo, Ads, and more. Outside of work, Binbin is passionate about bodybuilding and snowboarding. He is a NASM-certified personal trainer and an AASI-certified snowboard coach.