Tech Buzz China · Researcher profileandChina Research CollectiveAI ProemChina Research CollectiveAs of May 29, 2026
SONG Junxiao (宋俊潇)
Principal Researcher, DeepSeek
Source check
Known for
DeepSeek-R1 (co-author, published in Nature) · DeepSeek-Prover series
Current org
Schools
1
Articles
0
Videos
2
Links
Career path
Studied
→
Inventor of GRPO (the RL algorithm behind R1)
Zhejiang U BS, HKUST PhD (Daniel Palomar)
Profile
Source checkSONG Junxiao (宋俊潇) is a Principal Researcher at DeepSeek. He is a co-author of the landmark DeepSeek-R1 paper published in Nature, which demonstrated how reinforcement learning can incentivize reasoning capabilities in large language models.
Song was previously a Ph.D. researcher at the Hong Kong University of Science and Technology (HKUST), in the Department of Electronic and Computer Engineering. His research at DeepSeek focuses on reinforcement learning for reasoning, mathematical reasoning in LLMs, and automated theorem proving.
He has contributed to multiple DeepSeek model releases including DeepSeek-R1, DeepSeek-Prover, and DeepSeek-Prover-V2.
Known for
DeepSeek-R1 (co-author, published in Nature)DeepSeek-Prover seriesDeepSeek-Prover-V2
Education
Hong Kong University of Science and Technology
Ph.D. in Electronic and Computer Engineering
Articles / interviews
Profile links