PeopleofSuperintelligence

Researcher

Qinyu Chen

Qinyu Chen is a DeepSeek-AI researcher whose work spans mixture-of-experts language models, reasoning, conditional memory, and inference. In 2024, they coauthored DeepSeek-V2 and DeepSeek-V3; in 2025, they coauthored DeepSeek-R1 and DeepSeek-V3.2. Their 2026 work includes Engram’s conditional memory, DSpark’s speculative decoding, and DeepSeek-V4.1-Flash’s long-context model design. As of October 2026, they are part of DeepSeek-AI’s Research & Engineering team.

Now at
DeepSeek
Works
13
Citations
1,701
h-index
5

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. May 2024

    DeepSeek-V2 technical report

    In May 2024, they coauthor DeepSeek-V2, a mixture-of-experts language model with 236B total parameters and 21B activated for each token.

  2. December 2024

    DeepSeek-V3 technical report

    The December 2024 DeepSeek-V3 technical report describes a mixture-of-experts model with 671B total parameters and 37B activated for each token; Chen is a coauthor.

  3. September 2025

    DeepSeek-R1 in Nature

    In September 2025, they coauthor the Nature paper on DeepSeek-R1, which shows that pure reinforcement learning can encourage reasoning without human-labelled reasoning trajectories.

  4. December 2025

    DeepSeek-V3.2 technical report

    DeepSeek-V3.2 combines efficient attention, scalable reinforcement learning, and a pipeline for agent-training data; Chen coauthors the report in December 2025.

  5. July 2026

    Conditional memory with Engram

    In July 2026, they coauthor an ACL paper that introduces Engram, a conditional-memory module using local n-gram lookup. The paper reports gains on knowledge and reasoning benchmarks.

  6. July 2026

    DSpark speculative decoding

    They coauthor DSpark in July 2026, a speculative-decoding framework that combines parallel generation with adaptive, load-aware verification. In DeepSeek-V4 serving, it accelerates per-user generation by 60–85% at matched throughput.

  7. September 2026

    DeepSeek-V4.1-Flash model

    In September 2026, they coauthor DeepSeek-V4.1-Flash, a multimodal mixture-of-experts model with 552B backbone parameters and contexts up to one million tokens. Its design reduces the global key-value cache footprint to about one quarter of DeepSeek-V4-Flash.

  8. As of October 2026

    Research and engineering at DeepSeek-AI

    As of October 2026, they are part of DeepSeek-AI’s Research & Engineering team.

Most-cited works

  1. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning2025 · Nature · 1,118 citationsWith Bingxuan Wang, Chengqi Deng, Damai Dai, Daya Guo, Erhang Li and Fangyun Lin
  2. Can Open Large Language Models Catch Vulnerabilities?2025 · 539 citationsWith Bingxuan Wang, Chengqi Deng, Damai Dai, Erhang Li, Fangyun Lin and Haocheng Wang
  3. Robust pavement crack segmentation network based on transformer and dual-branch decoder2024 · Construction and Building Materials · 19 citations
  4. Managing the value Co-creation of peer service providers in the sharing economy: The perspective of customer incivility2023 · Heliyon · 13 citations
  5. Conflict with children, psychological depression, and problematic internet use among Chinese older adults: The moderating effect of sociability and living situation2023 · Digital Health · 11 citations

From OpenAlex, as of 21 October 2026.

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

  1. DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 21 October 2026.
  2. DeepSeek-V3 Technical Report arxiv.org. Checked 21 October 2026.
  3. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning nature.com. Checked 21 October 2026.
  4. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 21 October 2026.
  5. Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models aclanthology.org. Checked 21 October 2026.
  6. Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models arxiv.org. Checked 21 October 2026.
  7. DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation arxiv.org. Checked 21 October 2026.
  8. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 21 October 2026.