AI People Wiki

Builders & open source

Jason Wei

Works at Meta Superintelligence Labs, Meta

Jason Wei works at Meta Superintelligence Labs; from 2023 to 2025 Wei was at OpenAI, working on reasoning and agents, and before that Wei was a research scientist at Google Brain. Wei is first author of Chain-of-Thought Prompting Elicits Reasoning in Large Language Models and of Finetuned Language Models Are Zero-Shot Learners, which introduced the instruction-tuned FLAN model, and a co-author of Scaling Instruction-Finetuned Language Models. Wei's homepage also lists a 2024 paper on measuring short-form factuality in large language models and a 2025 benchmark for browsing agents.1234

At a glance

Current role
Works at Meta Superintelligence Labs
Company
Meta
Sector
Company
Work listed
5
Sources cited
8
Profile reviewed

Career

Jason Wei joined Google Brain in October 2020, then worked there as a research engineer from December 2021 and as a research scientist from June 2022, becoming a senior research scientist in October 2022. In February 2023 Wei joined OpenAI as a research scientist.5

Wei's homepage describes the OpenAI years, from 2023 to 2025, as work on reasoning and agents, and says Wei now works at Meta Superintelligence Labs. While at OpenAI, Wei appeared in company videos introducing o1 in ChatGPT and gave a keynote at OpenAI DevDay in 2024.15

Research

At Google, Wei was first author of Finetuned Language Models Are Zero-Shot Learners (2021), which introduced the instruction-tuned FLAN model, and of the 2022 chain-of-thought prompting paper, which showed that a few worked reasoning examples in a prompt improve performance on arithmetic, commonsense and symbolic reasoning tasks. Wei was also first author of Emergent Abilities of Large Language Models (2022) and a co-author of the follow-up on scaling instruction finetuning.3264

At OpenAI, Wei was co-first author of two evaluation benchmark papers: SimpleQA, introduced in a 2024 paper on measuring short-form factuality, and BrowseComp, a 2025 benchmark for browsing agents. Wei's CV also lists co-authorship of the PaLM paper and of a 2023 Nature paper on large language models and clinical knowledge.785

Teaching and talks

Wei has given guest lectures in university courses, including Stanford CS25 in 2023 and 2024, Stanford CS330, Harvard CS249r and MIT MAS.S68. Wei's homepage also lists a 2025 talk at the Stanford AI Club and a keynote at Columbia University's DAPLab.15

Work and contributions

  • Chain-of-thought prompting Published research

    First author of the 2022 paper showing that providing a few chain-of-thought demonstrations as exemplars improves large language models on arithmetic, commonsense, and symbolic reasoning tasks.2

  • FLAN instruction tuning Published research

    First author of Finetuned Language Models Are Zero-Shot Learners (2021), which instruction-tunes a 137B parameter model on over 60 NLP tasks and evaluates the resulting FLAN model on unseen task types; also a co-author of the 2022 follow-up on scaling instruction finetuning.34

  • Emergent Abilities of Large Language Models Published research

    First author of the 2022 paper that discusses abilities present in larger language models but not in smaller ones, written with co-authors including Yi Tay, Rishi Bommasani, Percy Liang and Jeff Dean.65

  • SimpleQA Published research

    First author, with Karina Nguyen, of the 2024 paper Measuring short-form factuality in large language models, which introduced the SimpleQA benchmark.75

  • BrowseComp Published research

    First author, with Zhiqing Sun, of BrowseComp, a 2025 benchmark for evaluating browsing agents.85

Timeline

  1. Published Easy Data Augmentation techniques for text classification with K. Zou at EMNLP 2019.5
  2. Joined Google Brain.5
  3. Released Finetuned Language Models Are Zero-Shot Learners, introducing FLAN.3
  4. Released the chain-of-thought prompting paper as first author.2
  5. Became a research scientist at Google Brain.5
  6. Released Emergent Abilities of Large Language Models as first author.6
  7. Became a senior research scientist at Google Brain.5
  8. Joined OpenAI as a research scientist.5
  9. Released the SimpleQA paper on measuring short-form factuality in language models.7
  10. Released BrowseComp, a benchmark for browsing agents, as co-first author.8

Other documented roles

Roles the sources document that are not confirmed as current.

Where they work

  • Meta 106 people in the wiki

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

8 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Jason Wei — homepageJason Wei · Official page · accessed Oct 21, 2026
  2. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models (arXiv 2201.11903)arXiv · Paper · published Jan 28, 2022 · accessed Oct 21, 2026
  3. Finetuned Language Models Are Zero-Shot Learners (arXiv 2109.01652)arXiv · Paper · published Sep 3, 2021 · accessed Oct 21, 2026
  4. Scaling Instruction-Finetuned Language Models (arXiv 2210.11416)arXiv · Paper · published Oct 20, 2022 · accessed Oct 21, 2026
  5. Jason Wei — CV (PDF)Jason Wei · Official page · accessed Oct 21, 2026
  6. Emergent Abilities of Large Language Models (arXiv 2206.07682)arXiv · Paper · published Jun 15, 2022 · accessed Oct 21, 2026
  7. Measuring short-form factuality in large language models (arXiv 2411.04368)arXiv · Paper · published Nov 7, 2024 · accessed Oct 21, 2026
  8. BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents (arXiv 2504.12516)arXiv · Paper · published Apr 16, 2025 · accessed Oct 21, 2026