SHAO Zhihong (邵智宏)

SHAO Zhihong (邵智宏) is a Research Scientist at DeepSeek, recognized as one of MIT Technology Review's 35 Innovators Under 35. He received his Ph.D. in Computer Science from Tsinghua University, advised by Professor Minlie Huang.

Shao's research focuses on LLM reasoning, tool augmentation, and mathematical problem solving. He is a key contributor to multiple landmark DeepSeek projects: DeepSeekMath (first author), DeepSeek-R1 (published in Nature), DeepSeek-Prover series, DeepSeek-Coder-V2, and DeepSeek-V3.

He also led the ToRA project (tool-integrated reasoning agent), which was the first open-source model to exceed 50% accuracy on the competition-level MATH dataset. His work spans informal math pre-training, formal theorem proving, and inference-time optimization for LLMs.