PeopleofSuperintelligence

Researcher

Dejian Yang

Dejian Yang is an AI researcher whose work spans chatbot dialogue, semantic parsing and large language models. He coauthored a dynamic-vocabulary chatbot paper in 2018 and a study of pretrained models' grounding abilities in 2021. His 2024 papers include DeepSeek-Coder, DeepSeek-V2, DeepSeek-Prover-V1.5 and DeepSeek-V3; he also coauthored the DeepSeek-R1 paper on reinforcement learning in 2025. As of October 2026, he is a researcher at DeepSeek. In September 2026, he coauthored DeepSeek Elastic Compute (DSec), a sandbox platform for agent training and evaluation.

Now at
DeepSeek
Works
11
Citations
541
h-index
6
Badges

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2018

    Dynamic vocabularies for chatbots

    At Beihang University, he coauthors the 2018 AAAI paper "Neural Response Generation With Dynamic Vocabularies." Its sequence-to-sequence model gives each input its own decoding vocabulary for chatbot responses.

  2. 2021

    Grounding pretrained language models

    For a 2021 ACL Findings paper, he studies how pretrained language models ground tokens to concepts. The paper pairs an erasing-then-awakening method with text-to-SQL parsers and reports up to a 9.8% absolute improvement.

  3. 2024

    DeepSeek-Coder for programming

    He coauthors the 2024 DeepSeek-Coder report, which introduces open-source code models from 1.3B to 33B parameters trained from scratch on two trillion tokens.

  4. 2024

    DeepSeek-V2's efficient architecture

    In 2024, he coauthors DeepSeek-V2, a 236B-parameter Mixture-of-Experts model with 21B parameters activated for each token and a 128K-token context window.

  5. 2024

    Lean theorem proving with Prover-V1.5

    The 2024 DeepSeek-Prover-V1.5 report lists him among its authors and describes a Lean 4 theorem-proving model refined with proof-assistant feedback and Monte Carlo tree search.

  6. December 2024

    DeepSeek-V3's 671B model

    In December 2024, he coauthors the DeepSeek-V3 technical report, which presents a Mixture-of-Experts model with 671B total parameters and 37B activated for each token.

  7. September 2025

    DeepSeek-R1 reasoning research

    Nature publishes the DeepSeek-R1 paper in September 2025 with him as a coauthor. It shows that pure reinforcement learning can produce reasoning patterns such as self-reflection, verification and dynamic strategy adaptation in language models.

  8. As of October 2026

    Researcher at DeepSeek

    As of October 2026, he is a core researcher at DeepSeek. In September 2026, he coauthors DeepSeek Elastic Compute (DSec), a production sandbox platform for training and evaluating language-model agents.

Career

  1. NowDeepSeekCore researcher
  2. 2017–2019Beihang University

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. DeepSeek-V3 Technical Report2024 · 292 citationsWith Bingxuan Wang, Bochao Wu, Chengqi Deng, Damai Dai and Daya Guo
  2. PPNE: Property Preserving Network Embedding2017 · Lecture notes in computer science · 77 citations
  3. Neural Response Generation With Dynamic Vocabularies2018 · Proceedings of the AAAI Conference on Artificial Intelligence · 70 citations
  4. From Properties to Links2017 · 53 citations
  5. Awakening Latent Grounding from Pretrained Language Models for Semantic Parsing2021 · 29 citations

From OpenAlex, as of 19 October 2026.

  1. The median h-index is 53 at universities and 19 at frontier AI labsIncludes Dejian Yang. The median h-index is 53 at universities and 19 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. The company share of AI researchers’ top papers rose from 2% to 32%Includes DeepSeek. We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 2% in 1975–1979 to 32% in 2020–2024.
  3. Google has the highest combined h-index among AI organizationsIncludes DeepSeek. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  4. Google alumni founded or led 87 other organizations, the most among the labs we compareIncludes DeepSeek. We trace 212 organizations founded or led by AI lab alumni, and 223 career moves between the labs.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Neural Response Generation With Dynamic Vocabularies ojs.aaai.org. Checked 19 October 2026.
  2. Awakening Latent Grounding from Pretrained Language Models for Semantic Parsing - Microsoft Research microsoft.com. Checked 19 October 2026.
  3. DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence arxiv.org. Checked 19 October 2026.
  4. DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 19 October 2026.
  5. DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search arxiv.org. Checked 19 October 2026.
  6. DeepSeek-V3 Technical Report arxiv.org. Checked 19 October 2026.
  7. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning nature.com. Checked 19 October 2026.
  8. YANG Dejian (杨德建) — China AI Atlas ai.techbuzzchina.com. Checked 19 October 2026.
  9. DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale arxiv.org. Checked 19 October 2026.