PeopleofSuperintelligence

Researcher

Jingyang Yuan

Also known as 袁境阳

Jingyang Yuan is a doctoral student at Peking University whose work on sparse attention helps language models process long contexts with less computation. During a DeepSeek internship, he contributes to DeepSeek-V2 and DeepSeek-V3 and coauthors DeepSeek-R1; his first-author Native Sparse Attention paper wins a Best Paper award at ACL 2025. He later coauthors the DeepSeek-V3.2, V4 and V4.1-Flash reports. As of October 2026, he is a doctoral student in Peking University's DLIB laboratory.

Badges

Timeline

  1. 2024–2025

    DeepSeek research internship

    He interns at DeepSeek and becomes a core contributor to DeepSeek-V2, whose report appears in May 2024, and DeepSeek-V3, whose report appears in December 2024. V2 has 236 billion total parameters and combines Multi-head Latent Attention (MLA) with DeepSeekMoE; V3 has 671 billion total parameters, trains on 14.8 trillion tokens and uses auxiliary-loss-free load balancing.

  2. January 2025–September 2025

    Reasoning with DeepSeek-R1

    He coauthors DeepSeek-R1 in January 2025. The work shows that pure reinforcement learning can develop reasoning in language models without human-labeled reasoning trajectories, and Nature publishes it in September 2025.

  3. February 2025–August 2025

    Native Sparse Attention

    In February 2025, he publishes Native Sparse Attention (NSA) as first author with colleagues at DeepSeek, Peking University and the University of Washington; DeepSeek's Wenfeng Liang is one of its corresponding authors. NSA combines coarse-grained token compression with fine-grained token selection and speeds up attention on 64k-length sequences; the paper appears in the proceedings of the 63rd ACL Annual Meeting. At ACL 2025 in Vienna, held from July to August 2025, he presents sparse attention as an important route for future long-text processing, and NSA wins one of the conference's four Best Paper awards.

  4. August 2025

    Doctoral study at Peking

    In August 2025, he moves from master's study to doctoral study in Peking University's School of Computer Science. Professor Ming Zhang, an NSA co-corresponding author, advises him.

  5. December 2025

    DeepSeek-V3.2 and sparse attention

    He coauthors DeepSeek-V3.2 in December 2025, which introduces DeepSeek Sparse Attention (DSA) to reduce long-context computation while preserving model performance.

  6. April 2026

    DeepSeek-V4 and million-token context

    He coauthors the DeepSeek-V4 report in April 2026, which presents Mixture-of-Experts models that support one-million-token contexts and a hybrid attention architecture combining Compressed Sparse Attention with Heavily Compressed Attention.

  7. September 2026

    The DeepSeek-V4.1-Flash report

    In September 2026, he coauthors DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model with 552 billion backbone parameters and support for contexts of up to one million tokens.

  8. As of October 2026

    Doctoral study in the DLIB lab

    As of October 2026, he is a doctoral student in Peking University's DLIB laboratory.

Career

  1. From 2024DeepSeekIntern

Education

  1. From 2025Peking UniversityDoctoral study

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Jingyang Yuan. 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 with dated positions at the same organization in the same years.

Sources and credits

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

  1. 计算机学院张铭教授团队与DeepSeek合作的论文获得ACL 2025最佳论文奖-北京大学计算机学院 cs.pku.edu.cn. Checked 21 October 2026.
  2. [2405.04434] DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 21 October 2026.
  3. [2412.19437] DeepSeek-V3 Technical Report arxiv.org. Checked 21 October 2026.
  4. [2501.12948] DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 21 October 2026.
  5. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning | Nature nature.com. Checked 21 October 2026.
  6. [2502.11089] Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention arxiv.org. Checked 21 October 2026.
  7. Jingyang Yuan - ACL Anthology aclanthology.org. Checked 21 October 2026.
  8. Awards - ACL 2025 2025.aclweb.org. Checked 21 October 2026.
  9. DeepSeek's Liang Wenfeng & Peking University's Yang Yaodong's Team Win Best Paper Award at ACL 2025 with NSA Paper eu.36kr.com. Checked 21 October 2026.
  10. Liang Wenfeng's Talent Philosophy Behind Huawei "Genius Young Man" Conflict with DeepSeek eu.36kr.com. Checked 21 October 2026.
  11. [2512.02556] DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 21 October 2026.
  12. [2606.19348] DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 21 October 2026.
  13. [2609.19969] DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 21 October 2026.
  14. 北京大学数据科学与工程所DLIB实验室 pkudlib.github.io. Checked 21 October 2026.