PeopleofSuperintelligence

Researcher

Xiaokang Zhang

Xiaokang Zhang is an AI researcher with the DeepSeek-AI Team in Hangzhou whose work spans mixture-of-experts language models, mathematical reasoning and AI infrastructure. He coauthored the 671-billion-parameter DeepSeek-V3 technical report in December 2024 and the DeepSeek-R1 reinforcement-learning study in 2025. His 2025 research also included verifier-guided theorem proving in DeepSeekMath-V2 and DeepSeek Sparse Attention in the December 2025 DeepSeek-V3.2 report. In 2026, he coauthored the April 2026 DeepSeek-V4 million-token context preview and the July 2026 DSpark speculative-decoding framework, and as of October 2026 he is a coauthor of the DeepSeek Elastic Compute sandbox platform for agentic training.

Works
8
Citations
1,985
h-index
5

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. December 2024

    DeepSeek-V3 technical report

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

  2. 2025

    DeepSeek-R1 and reinforcement learning

    With the DeepSeek-AI Team in Hangzhou, he coauthors the DeepSeek-R1 study on arXiv in January 2025 and in Nature in September 2025. The study shows that pure reinforcement learning can incentivize reasoning in large language models without human-labelled trajectories and facilitate patterns such as self-reflection, verification and dynamic strategy adaptation.

  3. 2025

    DeepSeekMath-V2 theorem proving

    His 2025 DeepSeekMath-V2 paper trains a mathematical proof generator using a verifier as its reward model. The generator is incentivized to identify and resolve issues in its own proofs before finalizing them.

  4. December 2025

    DeepSeek-V3.2 sparse attention

    He coauthors the DeepSeek-V3.2 report in December 2025, introducing DeepSeek Sparse Attention to reduce computational complexity while preserving model performance in long-context scenarios.

  5. April 2026

    DeepSeek-V4 million-token context models

    In April 2026, he coauthors the preview report for the DeepSeek-V4 series, which presents two mixture-of-experts language models for million-token context intelligence.

  6. July 2026

    DSpark speculative decoding

    The DSpark paper, which he coauthors in July 2026, introduces a speculative-decoding framework that unifies high-throughput parallel generation with adaptive, load-aware verification.

  7. As of October 2026

    DeepSeek Elastic Compute

    As of October 2026, he is a coauthor of DeepSeek Elastic Compute (DSec), a production sandbox infrastructure for agentic training submitted in September 2026. A single production-scale unit of DSec spans around 160 nodes, serving about 3 million sandboxes per day, supporting over 380,000 concurrent sandboxes and sustaining more than 5,000 sandbox creations per second.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. The United States employs an estimated 48% of people shaping AIWe estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.
  3. 43% of living people shaping AI work outside their country of birthWe estimate that 43% of living people shaping AI work outside their country of birth (520 of 1,198 people with known birth and employment countries).
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionWe estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

Sources and credits

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

  1. DeepSeek-V3 Technical Report arxiv.org. Checked 22 October 2026.
  2. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 22 October 2026.
  3. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning nature.com. Checked 22 October 2026.
  4. DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning github.com. Checked 22 October 2026.
  5. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 22 October 2026.
  6. DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 22 October 2026.
  7. DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation arxiv.org. Checked 22 October 2026.
  8. DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale arxiv.org. Checked 22 October 2026.