PeopleofSuperintelligence

Researcher

Zhibin Gou

Also known as 苟志斌

Zhibin Gou is a researcher at DeepSeek who works on language-model reasoning and reinforcement learning. He studied computer science at Beijing University of Posts and Telecommunications and completed a master's in artificial intelligence at Tsinghua University in 2025. At Microsoft Research Asia he coauthored ToRA, a tool-integrated reasoning agent for mathematics, and Rho-1, which uses Selective Language Modeling and whose paper was a NeurIPS 2024 Best Paper runner-up. He joined DeepSeek in June 2024, and his work there includes the DeepSeek-V3 technical report, the Nature 2025 DeepSeek-R1 paper, and the April 2026 DeepSeek-V4 report.

Now at
DeepSeek
Works
26
Citations
2,287
h-index
12

Publication numbers from OpenAlex, as of 17 October 2026.

Timeline

  1. 2018–2022

    Computer science at BUPT

    He completes a B.Sc. in computer science at Beijing University of Posts and Telecommunications from 2018 to 2022. The Ministry of Education of China awards him National Scholarships in 2018, 2019, and 2020, and in 2022 he receives the Outstanding Graduate and Outstanding Graduate Thesis awards in Beijing.

  2. September 2021–May 2022

    Dialogue research at Baidu

    From September 2021 to May 2022 he interns in the General Dialogue Group at Baidu in Beijing, working on open-domain dialogue and long-term memory. In 2022 he co-authors “Long Time No See! Open-Domain Conversation with Long-Term Persona Memory” in Findings of ACL 2022, which introduces the DuLeMon dataset and the PLATO-LTM framework.

  3. August 2022–June 2025

    A master's at Tsinghua

    He completes an M.Sc. in artificial intelligence at Tsinghua University, in the Division of Information Science and Technology at SIGS in Beijing, from August 2022 to June 2025. Yujiu Yang advises him.

  4. January 2023–May 2024

    Internship at Microsoft Research

    From January 2023 to May 2024 he interns in the NLC Group at Microsoft Research Asia in Beijing, with mentors Yeyun Gong and Weizhu Chen, on language-model reasoning and tool use. At ICLR 2024 he and Zhihong Shao are equal-contribution authors of ToRA, a tool-integrated reasoning agent for mathematical problem solving, with affiliations at Tsinghua University and Microsoft.

  5. 2024

    A NeurIPS runner-up

    With Zhenghao Lin, he is an author of “Not All Tokens Are What You Need for Pretraining,” a NeurIPS 2024 Best Paper runner-up. The paper introduces Rho-1, which uses Selective Language Modeling to train on selected tokens.

  6. Since June 2024

    Research at DeepSeek

    Since June 2024 he works as a researcher at DeepSeek in Beijing on large language models, reasoning, and reinforcement learning. In December 2024 he is an author of the DeepSeek-V3 technical report, which describes a mixture-of-experts language model with 671 billion total parameters and 37 billion activated for each token. In September 2025 Nature publishes the DeepSeek-R1 paper, with him as a core contributor on the DeepSeek-AI team in Hangzhou, and in April 2026 he is an author of the DeepSeek-V4 report on models with a one-million-token context.

Sources and credits

Photo: We have no published photo of Zhibin Gou yet. Send us a free-licensed photo.

  1. Zhibin Gou 苟志斌 - Zhibin Gou zubingou.github.io. Checked 16 October 2026.
  2. Zhibin Gou - ACL Anthology aclanthology.org. Checked 16 October 2026.
  3. ToRA: Tool-Integrated Reasoning Agents microsoft.github.io. Checked 16 October 2026.
  4. NeurIPS 2024 Awards neurips.cc. Checked 16 October 2026.
  5. [2412.19437] DeepSeek-V3 Technical Report arxiv.org. Checked 16 October 2026.
  6. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning | Nature nature.com. Checked 16 October 2026.
  7. [2606.19348] DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 16 October 2026.