PeopleofSuperintelligence

Researcher

Kai Dang

Kai Dang is an AI researcher who co-authored technical reports for the Qwen large language model series, from the first Qwen report in September 2023 to Qwen3 in May 2025. After undergraduate and master's study at Nankai University, he joined Alibaba Group as a researcher in 2022 and co-authored OFASys, a system for building multimodal generalist models. His later papers focused on reinforcement learning for language models: Group Sequence Policy Optimization (GSPO), released in July 2025, stabilized mixture-of-experts RL training and contributed to the Qwen3 models, and a December 2025 study shared recipes for stable RL training. In August 2026, he co-authored Qwen-CUA, an agent that uses computers through screenshots, keyboard and mouse.

Now at
Alibaba
Badges

Timeline

  1. 2015–2022

    Studies at Nankai University

    He studies at Nankai University as an undergraduate from 2015 to 2019 and as a master's student from 2019 to 2022.

  2. 2022

    OFASys at Alibaba Group

    He joins Alibaba Group as a researcher in 2022. In December 2022, he co-authors OFASys, a system for building generalist models; its single OFA+ model handles text, image, speech, video and motion data.

  3. September 2023

    The first Qwen technical report

    In September 2023, he co-authors the Qwen Technical Report, which introduces Qwen as the first installment of a large language model series. The series includes base pretrained models and Qwen-Chat chat models.

  4. December 2024–February 2025

    Qwen2.5 and Qwen2.5-VL

    In December 2024, he co-authors the Qwen2.5 Technical Report, which scales pretraining data from 7 trillion to 18 trillion tokens. In February 2025, he co-authors the Qwen2.5-VL Technical Report on the flagship model of the Qwen vision-language series.

  5. May 2025

    The Qwen3 technical report

    In May 2025, he co-authors the Qwen3 Technical Report. Qwen3 models range from 0.6 to 235 billion parameters and combine thinking and non-thinking modes in one framework.

  6. July 2025

    Group Sequence Policy Optimization

    In July 2025, he co-authors Group Sequence Policy Optimization (GSPO) with Chujie Zheng and colleagues. The reinforcement learning algorithm stabilizes mixture-of-experts RL training and contributes to improvements in the Qwen3 models.

  7. December 2025

    Recipes for stable RL training

    In December 2025, he is second author, after Chujie Zheng, of "Stabilizing Reinforcement Learning with LLMs: Formulation and Practices". The paper tests RL training on a 30B mixture-of-experts model over hundreds of thousands of GPU hours and shares recipes for stable RL training.

  8. August 2026

    The Qwen-CUA computer-use agent

    In August 2026, he co-authors Qwen-CUA, a native computer-use agent built on a 397B-A17B Qwen mixture-of-experts backbone. The agent observes only screenshots and acts through keyboard and mouse events.

Career

  1. From 2022AlibabaResearcher

Education

  1. 2022Nankai UniversityMaster's student
  2. 2019Nankai UniversityUndergraduate student

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Kai Dang. 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.
  2. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Alibaba. 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.
  3. Universities (31%) and startups and other companies (25%) employ the largest shares of 2020–2026 entrantsIncludes Alibaba. Among 819 people whose first recorded role started in 2020–2026 and who have a known current organization, 60% work in commercial sectors (25% at startups and other companies, 23% at frontier labs and 13% at big tech, with 9 in two), and 31% work at universities, compared with 66% for earlier entrants.
  4. The top 10 organizations employed 22% of people shaping AI in 2025: 51% of company researchers and 21% of university researchersIncludes Alibaba. We estimate the share of people shaping AI who worked at the top 10 organizations each year from 1955 to 2025, comparing concentration across companies, universities, and institutes and labs.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Kai Dang — Lacuna lacuna.tiptreesystems.com. Checked 20 October 2026.
  2. [2212.04408] OFASys: A Multi-Modal Multi-Task Learning System for Building Generalist Models arxiv.org. Checked 20 October 2026.
  3. [2309.16609] Qwen Technical Report arxiv.org. Checked 20 October 2026.
  4. [2412.15115] Qwen2.5 Technical Report arxiv.org. Checked 20 October 2026.
  5. [2502.13923] Qwen2.5-VL Technical Report arxiv.org. Checked 20 October 2026.
  6. [2505.09388] Qwen3 Technical Report arxiv.org. Checked 20 October 2026.
  7. [2507.18071] Group Sequence Policy Optimization arxiv.org. Checked 20 October 2026.
  8. Chujie Zheng - Chujie Zheng 郑楚杰 chujiezheng.github.io. Checked 20 October 2026.
  9. [2512.01374] Stabilizing Reinforcement Learning with LLMs: Formulation and Practices arxiv.org. Checked 20 October 2026.
  10. [2608.02352] Qwen-CUA: Native Computer Use for (almost) Everything arxiv.org. Checked 20 October 2026.