PeopleofSuperintelligence

Researcher

Pavel Izmailov

Pavel Izmailov is a computer scientist whose research covers generalization in deep learning, reasoning in language models and the alignment of future superhuman AI. He studied at Lomonosov Moscow State University and Cornell before earning a computer science PhD at NYU in 2023 under Andrew Gordon Wilson. In 2018 he coauthored the Stochastic Weight Averaging paper, and a 2022 paper on the marginal likelihood won an ICML Outstanding Paper Award. At OpenAI he worked on weak-to-strong generalization with the superalignment team in 2023 and contributed to the o1 reasoning model in 2024. He is an assistant professor at NYU Tandon and a researcher at Anthropic, where he contributed to Claude 3.7 and Claude 4.

Now at
New York University Tandon School of Engineering, Courant Institute of Mathematical Sciences, New York University and Anthropic
Badges

Timeline

  1. 2017–2019

    A PhD start at Cornell

    He earns a BSc in applied math and computer science at Lomonosov Moscow State University. From 2017 to 2019, he is a PhD student in Operations Research and Information Engineering at Cornell, then receives an MSc and transfers to NYU.

  2. 2018

    Stochastic Weight Averaging

    In 2018, he coauthors "Averaging Weights Leads to Wider Optima and Better Generalization" with Dmitrii Podoprikhin, Timur Garipov, Dmitry Vetrov and Andrew Gordon Wilson. The paper shows that simple averaging of multiple points along the SGD trajectory, called Stochastic Weight Averaging (SWA), leads to better generalization than conventional training.

  3. 2022

    ICML Outstanding Paper Award

    In 2022, "Bayesian Model Selection, the Marginal Likelihood, and Generalization", written with Sanae Lotfi, Gregory Benton, Micah Goldblum and Andrew Gordon Wilson, receives an Outstanding Paper Award at the International Conference on Machine Learning. The paper studies the marginal likelihood, or Bayesian evidence, as a way to compare hypotheses that are entirely consistent with observations.

  4. 2023

    A PhD at NYU

    He defends his PhD in computer science at NYU in 2023, supervised by Andrew Gordon Wilson.

  5. December 2023

    Weak-to-strong generalization at OpenAI

    In December 2023, he is a coauthor of "Weak-to-Strong Generalization", work he does on OpenAI's superalignment team under Jeff Wu, Jan Leike and Ilya Sutskever. The paper finds that strong pretrained models finetuned on labels from a weak model consistently perform better than their weak supervisors. WIRED and MIT Technology Review cover the result, and he calls it "a really surprising and positive result".

  6. 2024

    Contributing to OpenAI o1

    In 2024, he is a contributor to OpenAI's "Learning to Reason with LLMs" post on o1, a large language model trained with reinforcement learning to perform complex reasoning. The model produces a long internal chain of thought before responding to a query.

  7. June 2025

    A professorship at NYU Tandon

    In June 2025, NYU Tandon welcomes him as a new assistant professor in its Computer Science and Engineering department and the Courant Department of Computer Science. His work at Anthropic includes the development of Claude 3.7 and Claude 4.

  8. As of October 2026

    Professor and Anthropic researcher

    He is an assistant professor in the NYU Tandon Computer Science and Engineering department, with a courtesy appointment at Courant CS, and a researcher at Anthropic. His 2026 work includes "How's it going? Reinforcement learning in language models recruits a functional welfare axis" with David J. Chalmers and a NeurIPS 2026 paper on reasoning from pretraining to post-training.

Career

  1. NowNew York University Tandon School of EngineeringAssistant Professor
  2. NowCourant Institute of Mathematical Sciences, New York UniversityCourtesy Appointment
  3. NowAnthropicResearcher
  4. 2023–2024OpenAI

Education

  1. 2023New York UniversityPhD
  2. 2019Cornell UniversityPhD
  3. 2019Cornell UniversityMSc
  4. Lomonosov Moscow State UniversityBSc

From the biography.

  1. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Pavel Izmailov. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  2. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Pavel Izmailov. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  3. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Pavel Izmailov. 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.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Pavel Izmailov. 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.

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 Pavel Izmailov yet. Send us a free-licensed photo.

  1. Pavel Izmailov izmailovpavel.github.io. Checked 22 October 2026.
  2. Pavel Izmailov | NYU Tandon School of Engineering engineering.nyu.edu. Checked 22 October 2026.
  3. [1803.05407] Averaging Weights Leads to Wider Optima and Better Generalization arxiv.org. Checked 22 October 2026.
  4. ICML 2022 Awards icml.cc. Checked 22 October 2026.
  5. Bayesian Model Selection, the Marginal Likelihood, and Generalization proceedings.mlr.press. Checked 22 October 2026.
  6. NYU Tandon welcomes new faculty members pushing the boundaries of robotics and AI engineering.nyu.edu. Checked 22 October 2026.
  7. [2312.09390] Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision arxiv.org. Checked 22 October 2026.
  8. OpenAI’s Ilya Sutskever Has a Plan for Keeping Super-Intelligent AI in Check | WIRED wired.com. Checked 22 October 2026.
  9. Now we know what OpenAI’s superalignment team has been up to | MIT Technology Review technologyreview.com. Checked 22 October 2026.
  10. Pavel Izmailov izmailovpavel.github.io. Checked 22 October 2026.