PeopleofSuperintelligence

Researcher

Wangding Zeng

Also known as zwd003

Wangding Zeng is a researcher at DeepSeek-AI whose work spans expert specialization, language and code models, reasoning, and efficient attention. He began a master's degree at the AI Institute of Beijing University of Posts and Telecommunications in 2021. Across 2024 and 2025, he coauthored DeepSeekMoE, DeepSeek-V2 and DeepSeek-Coder-V2, DeepSeek-V3 and DeepSeek-R1, and Native Sparse Attention; that paper won an ACL 2025 Best Paper award. In July 2026, he coauthored Conditional Memory via Scalable Lookup, which uses local n-grams to retrieve from a large embedding table. As of October 2026, he works at DeepSeek-AI.

Now at
DeepSeek
Badges

Timeline

  1. 2021

    A master's at BUPT

    He begins a master's degree at the AI Institute of Beijing University of Posts and Telecommunications in 2021.

  2. 2024

    DeepSeekMoE architecture

    In 2024, he coauthors DeepSeekMoE, which divides experts into smaller units and isolates shared experts to capture common knowledge.

  3. May–June 2024

    DeepSeek-V2 and Coder-V2

    In May 2024, he coauthors DeepSeek-V2, whose multi-head latent attention (MLA) compresses the key-value cache into a latent vector; DeepSeek also credits him with key innovations to MLA. In June 2024, he coauthors DeepSeek-Coder-V2, an open-source code model further pretrained from an intermediate DeepSeek-V2 checkpoint with an additional 6 trillion tokens.

  4. December 2024

    DeepSeek-V3 technical report

    In December 2024, he coauthors the DeepSeek-V3 technical report, which describes a mixture-of-experts model with 671 billion parameters and 37 billion activated per token.

  5. January 2025

    Reinforcement learning for R1

    In January 2025, he coauthors DeepSeek-R1; DeepSeek-R1-Zero trains through reinforcement learning without supervised fine-tuning, and reasoning patterns such as reflection and verification emerge during training.

  6. July 2025

    Native Sparse Attention award

    In July 2025, he coauthors Native Sparse Attention, which combines token compression and token selection for long-context modeling. It wins an ACL 2025 Best Paper award.

  7. July 2026

    Conditional Memory via Scalable Lookup

    In July 2026, he coauthors Conditional Memory via Scalable Lookup, which uses local n-grams to index a large embedding table for retrieval.

  8. As of October 2026

    Research at DeepSeek-AI

    As of October 2026, he works at DeepSeek-AI.

Career

  1. NowDeepSeekResearcher

Education

  1. From 2021Beijing University of Posts and TelecommunicationsMaster's

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Wangding Zeng. 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. 16% of AI career moves in the 2020s went from a university to a companyIncludes Wangding Zeng. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  3. 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.
  4. 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.

Sources and credits

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

  1. China's AI disrupter DeepSeek bets on low-key team of 'young geniuses' to beat US giants koreatimes.co.kr. Checked 21 October 2026.
  2. DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models aclanthology.org. Checked 21 October 2026.
  3. DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 21 October 2026.
  4. DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence arxiv.org. Checked 21 October 2026.
  5. DeepSeek-V3 Technical Report arxiv.org. Checked 21 October 2026.
  6. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 21 October 2026.
  7. Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention aclanthology.org. Checked 21 October 2026.
  8. Awards - ACL 2025 2025.aclweb.org. Checked 21 October 2026.
  9. Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models aclanthology.org. Checked 21 October 2026.
  10. zwd003 (wangding zeng) · GitHub github.com. Checked 21 October 2026.