Xianjun Yang

I am a Member of Technical Staff at Reflection AI, working on open intelligence for LLMs. I completed my PhD in Computer Science at University of California, Santa Barbara, where I was co-advised by Professor Linda Petzold, William Yang Wang, and Stephen Wilson. Prior to joining Reflection AI, I worked at Meta Superintelligence Lab, where I work across FAIR, GenAI.

My recent research focuses on Agentic RL & Mid-training and previously I worked heavily on AI Safety, LLM Security and Alignment, Mechanistic Interpretability. I have published extensively at top-tier venues including ICML(Oral), NeurIPS, ICLR(Oral), and ACL, with around 3k citations.

Prior to AI, I quited my PhD in Fluid Mechanics when I leart AGI is coming. Then I earned my CS phd within 3 years without any prior cs background. I received my B.S. in Energy and Power Engineering from Tsinghua University.