Xiao Ma is an alignment researcher at Meta Superintelligence Labs, where she recently led alignment of the image and video planner agent. Her work spans post-training, reinforcement learning, and scalable safety evaluations. She was a critical contributor to the launch of Muse Image.

Previously a Staff Software Engineer at Google DeepMind, she co-invented and empirically validated RLGF (grading feedback), brought it into production to unblock Gemini 2.5 Pro and beyond, and scaled its adoption across Gemini. She also co-authored an ICLR 2025 Outstanding Paper on shallow safety alignment.

Building on her research in human–AI interaction and social trust, she is interested in security and social engineering risks in emerging hybrid human–agent networks, and in establishing protocols for trust within them.