PeopleofSuperintelligence

Researcher

Yunfan Xiong

Yunfan Xiong is a researcher whose work speeds up large language model inference. He earned a Bachelor of Science degree from Peking University's School of Electronics Engineering and Computer Science in 2022 and completed a data science master's at the university's Academy for Advanced Interdisciplinary Studies in 2025. He coauthored DySpec, DeepSeek-V3 and DeepSeek-R1. In 2026, he coauthored DSpark and DeepSeek-V4.1-Flash, which uses cross-layer KV cache reuse and FP4 caching to reduce its memory footprint. As of October 2026, he works at DeepSeek on large language model inference acceleration.

Now at
DeepSeek
Badges

Timeline

  1. 2022

    A Peking University degree

    He earns a Bachelor of Science degree from Peking University's School of Electronics Engineering and Computer Science in 2022.

  2. 2022–2025

    Data science master's

    He pursues a master's degree in Data Science at Peking University's Academy for Advanced Interdisciplinary Studies from 2022 to 2025.

  3. October 2024

    DySpec speculative decoding

    In October 2024, he coauthors DySpec, a speculative decoding algorithm with a dynamic token tree for faster language-model inference.

  4. December 2024

    The DeepSeek-V3 report

    In December 2024, he coauthors the DeepSeek-V3 Technical Report, which describes a 671B-parameter mixture-of-experts language model.

  5. September 2025

    DeepSeek-R1 reasoning paper

    In September 2025, he coauthors the DeepSeek-R1 paper. It shows that pure reinforcement learning can elicit reasoning in language models without human-labelled reasoning trajectories.

  6. July 2026

    The DSpark framework

    In July 2026, he coauthors DSpark, a speculative decoding framework with confidence-scheduled verification. The paper reports its deployment in DeepSeek-V4 serving.

  7. September 2026

    DeepSeek-V4.1-Flash model

    In September 2026, he coauthors DeepSeek-V4.1-Flash, a multimodal model that combines cross-layer KV cache reuse with FP4 caching to reduce its global KV cache footprint.

  8. As of October 2026

    Inference work at DeepSeek

    As of October 2026, he works at DeepSeek on large language model inference acceleration.

Career

  1. NowDeepSeek

Education

  1. 2025Peking UniversityMaster's in Data Science
  2. 2022Peking UniversityBachelor of Science

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Yunfan Xiong. 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 Yunfan Xiong. 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 Yunfan Xiong yet. Send us a free-licensed photo.

  1. 北京大学王选所数据管理实验室 mod.icst.pku.edu.cn. Checked 20 October 2026.
  2. DySpec: Faster Speculative Decoding with Dynamic Token Tree Structure arxiv.org. Checked 20 October 2026.
  3. DeepSeek-V3 Technical Report arxiv.org. Checked 20 October 2026.
  4. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning - PMC pmc.ncbi.nlm.nih.gov. Checked 20 October 2026.
  5. DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation arxiv.org. Checked 20 October 2026.
  6. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 20 October 2026.
  7. 北京大学王选所数据管理实验室 mod.wict.pku.edu.cn. Checked 20 October 2026.