PeopleofSuperintelligence

Researcher

Zongyu Lin

Also known as Johnson Lin

Zongyu Lin is an AI researcher whose work spans language models, visual reasoning, video generation and robot learning. They earned an electronic engineering bachelor's degree at Tsinghua University and began computer science PhD research at UCLA, co-advised by Yizhou Sun and Kai-Wei Chang. At Moonshot AI, they worked on long-context pretraining that led to Kimi Chat. Their research includes a 2023 zero-shot NLP discriminator, the 2024 VDebugger framework, 2025 STIV and DreamGen. Since 2025, they have worked at Meta, where they are a research scientist in the TBD lab; they served as a core member of pretraining and co-led a Muse Spark model series.

Now at
Meta
Badges

Timeline

  1. 2017–2021

    Electronic engineering at Tsinghua

    They study electronic engineering at Tsinghua University from 2017 to 2021 and earn a bachelor's degree.

  2. 2022–2023

    Kimi's long-context pretraining

    From 2022 to 2023, they work as a founding research scientist at Moonshot AI on LLM pretraining and long context. They help pretrain Kimi Chat.

  3. Since 2023

    Computer science PhD at UCLA

    They begin computer science PhD research at UCLA in 2023, co-advised by Yizhou Sun and Kai-Wei Chang. They are on leave from the program.

  4. 2023

    A universal NLP discriminator

    In 2023, they coauthor A Universal Discriminator for Zero-Shot Generalization, an ACL paper that trains one discriminator to outperform generative approaches on many NLP tasks.

  5. 2024

    Debugging visual programs

    In 2024, they coauthor VDebugger, a critic-refiner framework that tracks execution step by step to locate and debug errors in visual programs.

  6. August 2025

    Text and image video generation

    At Apple, they coauthor STIV, published in August 2025. The method combines text and image conditions for video generation.

  7. 2025

    DreamGen for robot learning

    In 2025, they coauthor DreamGen, a pipeline that uses synthetic robot data from video world models to train policies that generalize across behaviors and environments.

  8. Since 2025

    Research scientist at Meta

    They join Meta in 2025. They are a research scientist at Meta's TBD lab, serve as a core member of pretraining, and co-lead a Muse Spark model series.

Career

  1. Since 2025MetaResearch scientist
  2. 2025Apple
  3. 2022–2023Moonshot AIFounding research scientist

Education

  1. From 2023University of California, Los AngelesPhD
  2. 2021Tsinghua UniversityBachelor's degree

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Zongyu Lin. 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. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  3. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. 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.
  4. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Meta. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.

Colleagues

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

Sources and credits

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

  1. Zongyu Lin rafa-zy.github.io. Checked 21 October 2026.
  2. Zongyu Lin | alphaXiv alphaxiv.org. Checked 21 October 2026.
  3. A Universal Discriminator for Zero-Shot Generalization - ACL Anthology aclanthology.org. Checked 21 October 2026.
  4. [2406.13444] VDebugger: Harnessing Execution Feedback for Debugging Visual Programs arxiv.org. Checked 21 October 2026.
  5. STIV: Scalable Text and Image Conditioned Video Generation - Apple Machine Learning Research machinelearning.apple.com. Checked 21 October 2026.
  6. [2505.12705] DreamGen: Unlocking Generalization in Robot Learning through Video World Models arxiv.org. Checked 21 October 2026.