PeopleofSuperintelligence

Researcher

Zhihong Shao

Also known as 邵智宏

Zhihong Shao is an AI researcher whose work connects language models with mathematical reasoning, retrieval and tool use. His early text-generation research appeared in a 2019 paper, followed by retrieval-augmented language-model work in 2023. DeepSeekMath and ToRA shaped his 2024 work, and MIT Technology Review named him to its 2024 Innovators Under 35 list. DeepSeek-Prover, DeepSeek-V3, DeepSeek-R1 and DeepSeekMath-V2 extended his work on proof generation, reasoning and language models. He holds a computer science PhD from Tsinghua University, advised by Minlie Huang, and is a Research Scientist at DeepSeek as of October 2026.

Now at
DeepSeek

Timeline

  1. 2019

    Long-form text generation

    In 2019, he coauthors Long and Diverse Text Generation with Planning-based Hierarchical Variational Model, which uses a planning step to organize sentences into longer, more varied text.

  2. 2023

    Retrieval-augmented language models

    In 2023, Shao coauthors Iter-RetGen, a retrieval-augmented method that uses an initial model response to find more relevant information for a later response.

  3. 2024

    DeepSeekMath and ToRA

    In 2024, he coauthors DeepSeekMath, whose 7B model scores 51.7% on MATH and whose paper introduces GRPO for reinforcement learning. He also coauthors ToRA, an ICLR 2024 agent that combines language reasoning with external tools. MIT Technology Review names him to its 2024 Innovators Under 35 list; he is a first author of DeepSeekMath.

  4. 2024

    Formal theorem proving

    In 2024, he coauthors DeepSeek-Prover, which trains a theorem-proving model with synthetic Lean 4 data. DeepSeek-Prover-V1.5 adds reinforcement learning from proof-assistant feedback and a Monte Carlo tree search that explores proof paths.

  5. 2024

    DeepSeek-V3 engineering credits

    In December 2024, DeepSeek places him in the Research & Engineering credits for its DeepSeek-V3 technical report. The report describes a 671B-parameter mixture-of-experts model with 37B active parameters per token.

  6. 2025

    DeepSeek-R1 reasoning

    In 2025, he coauthors DeepSeek-R1, published in Nature. Its paper shows that reinforcement learning can elicit reasoning patterns in language models without human-labeled reasoning trajectories.

  7. November 2025

    A self-verifying math model

    In November 2025, he coauthors DeepSeekMath-V2, a model for self-verifying mathematical reasoning. Its paper reports gold-level results on IMO 2025 and CMO 2024, plus a 118/120 score on the 2024 Putnam.

  8. As of October 2026

    Research Scientist at DeepSeek

    As of October 2026, he is a Research Scientist at DeepSeek. He holds a computer science PhD from Tsinghua University, advised by Minlie Huang.

Career

  1. NowDeepSeekResearch Scientist

Education

  1. Tsinghua UniversityPhD

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes DeepSeek. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes DeepSeek. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staffIncludes DeepSeek. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staff. We test the estimate against 3,913 named model contributors and dated all-staff records across Google DeepMind, OpenAI and Anthropic, plus six additional frontier-grade labs.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes DeepSeek. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. [1908.06605] Long and Diverse Text Generation with Planning-based Hierarchical Variational Model arxiv.org. Checked 20 October 2026.
  2. [2305.15294] Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy arxiv.org. Checked 20 October 2026.
  3. [2402.03300] DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models arxiv.org. Checked 20 October 2026.
  4. [2309.17452] ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving arxiv.org. Checked 20 October 2026.
  5. Zhihong Shao | Innovators Under 35 innovatorsunder35.com. Checked 20 October 2026.
  6. [2405.14333] DeepSeek-Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data arxiv.org. Checked 20 October 2026.
  7. [2408.08152] DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search arxiv.org. Checked 20 October 2026.
  8. DeepSeek-V3 Technical Report arxiv.org. Checked 20 October 2026.
  9. [2412.19437] DeepSeek-V3 Technical Report arxiv.org. Checked 20 October 2026.
  10. [2501.12948] DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 20 October 2026.
  11. [2511.22570] DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning arxiv.org. Checked 20 October 2026.
  12. Zhihong Shao 邵智宏 - Zhihong Shao 邵智宏 zhihongshao.github.io. Checked 20 October 2026.