PeopleofSuperintelligence

Researcher

Pei Zhang

Pei Zhang is an AI researcher whose work spans machine translation, language models, and speech systems. He coauthored Alibaba's WMT18 translation submission, speech and document translation methods, PolyLM, and reports on Qwen2 and Qwen3. His 2025 and 2026 work included CultureSynth, PolyMath, Qwen3-Omni, Qwen3-ASR, and Qwen3-TTS. As of October 2026, he is part of Alibaba Cloud's Qwen Team.

Now at
Alibaba
Works
25

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2018

    Alibaba's WMT18 systems

    In 2018, he coauthors Alibaba's neural machine-translation systems for WMT18.

  2. 2019–2022

    Speech and document translation

    In 2019, he coauthors a lattice transformer that uses multiple ASR paths for speech translation. In 2020, he coauthors a long-short-term masking method for document-level translation. In 2022, he coauthors a competency-aware model that estimates translation quality.

  3. 2023

    PolyLM's multilingual models

    In 2023, he coauthors PolyLM, an open-source multilingual language model trained on 640 billion tokens and available in 1.7B and 13B sizes. Its curriculum increases the proportion of non-English data from 30 percent to 60 percent during pretraining.

  4. July 2024–December 2024

    Qwen2 and Qwen2.5

    In July 2024, he coauthors the Qwen2 report on models ranging from 0.5 to 72 billion parameters. In December 2024, he coauthors the Qwen2.5 report, which describes training data scaled to 18 trillion tokens and multistage reinforcement learning.

  5. 2025

    CultureSynth and PolyMath

    In 2025, he coauthors CultureSynth, a framework for generating cultural question-answer pairs from a multilingual taxonomy. He also coauthors PolyMath, a mathematical-reasoning benchmark across 18 languages and four difficulty levels with Alibaba's Qwen Team.

  6. 2025

    The Qwen3 model family

    He coauthors the 2025 Qwen3 report, which presents dense and mixture-of-experts models with parameter scales from 0.6 to 235 billion. Alibaba Cloud releases Qwen3 in April 2025.

  7. September 2025

    Qwen3-Omni in real time

    In September 2025, he coauthors Qwen3-Omni, an end-to-end model from Alibaba Cloud's Qwen team that handles text, images, audio, and video and generates speech in real time.

  8. As of October 2026

    Qwen Team speech systems

    In 2026, he coauthors Qwen3-ASR, which supports recognition in 52 languages and dialects, and Qwen3-TTS, a multilingual streaming text-to-speech family. As of October 2026, he is part of Alibaba Cloud's Qwen Team and contributes to Qwen3.5-Omni.

  1. 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.
  2. 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.
  3. 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.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Alibaba. 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.

Sources and credits

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

  1. Alibaba’s Neural Machine Translation Systems for WMT18 aclanthology.org. Checked 20 October 2026.
  2. Lattice Transformer for Speech Translation aclanthology.org. Checked 20 October 2026.
  3. Long-Short Term Masking Transformer: A Simple but Effective Baseline for Document-level Neural Machine Translation aclanthology.org. Checked 20 October 2026.
  4. Competency-Aware Neural Machine Translation: Can Machine Translation Know its Own Translation Quality? aclanthology.org. Checked 20 October 2026.
  5. PolyLM: An Open Source Polyglot Large Language Model arxiv.org. Checked 20 October 2026.
  6. Qwen2 Technical Report arxiv.org. Checked 20 October 2026.
  7. Qwen2.5 Technical Report arxiv.org. Checked 20 October 2026.
  8. CultureSynth: A Hierarchical Taxonomy-Guided and Retrieval-Augmented Framework for Cultural Question-Answer Synthesis aclanthology.org. Checked 20 October 2026.
  9. PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts qwen-polymath.github.io. Checked 20 October 2026.
  10. PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts arxiv.org. Checked 20 October 2026.
  11. Qwen3 Technical Report arxiv.org. Checked 20 October 2026.
  12. Qwen3 github.com. Checked 20 October 2026.
  13. Qwen3-Omni Technical Report arxiv.org. Checked 20 October 2026.
  14. Qwen3-Omni github.com. Checked 20 October 2026.
  15. Qwen3-ASR Technical Report arxiv.org. Checked 20 October 2026.
  16. Qwen3-TTS Technical Report arxiv.org. Checked 20 October 2026.
  17. Jialong Tang’s Homepage tangjialong.github.io. Checked 20 October 2026.