PeopleofSuperintelligence

Researcher

Hongseok Namkoong

Also known as Hong

Researcher Hongseok Namkoong builds trustworthy AI for real-world decision-making. In 2019, he was a Stanford PhD candidate, with an M.S. in statistics from Stanford and a B.S. in industrial engineering and mathematics from KAIST. He coauthored a 2017 NeurIPS best-paper work on variance regularization and a 2018 ICML runner-up paper on fairness without demographic labels. He later coauthored Model Soups at ICML in 2022, PersonalLLM at ICLR in 2025, and MBABench, a NeurIPS 2026 benchmark for LLM agents. Before joining Columbia, he spent a year as a research scientist on Meta's Adaptive Experimentation team; he is an associate professor in Columbia Business School's Decision, Risk, and Operations division and a member of the Columbia Data Science Institute.

Timeline

  1. 2017

    Variance-based regularization

    With John C. Duchi, he coauthors Variance-based Regularization with Convex Objectives in 2017. The paper develops a convex surrogate for variance in risk minimization and stochastic optimization, and its conference version wins the NeurIPS 2017 Best Paper Award.

  2. 2018

    Fairness without demographics

    In 2018, he coauthors Fairness Without Demographics in Repeated Loss Minimization with Tatsunori Hashimoto, Megha Srivastava, and Percy Liang. The method uses distributionally robust optimization to control minority-group risk at each step while remaining oblivious to group identity. The paper wins an ICML Best Paper Runner-up Award.

  3. February 2019

    A Stanford PhD candidate

    In February 2019, he presents work on learning models with uniform subpopulation performance as a PhD candidate in Stanford's Department of Management Science & Engineering, advised by John Duchi and Peter Glynn. He already holds an M.S. in statistics from Stanford and a B.S. in industrial engineering and mathematics from KAIST.

  4. 2022

    Model soups

    In 2022, he coauthors Model Soups at ICML, showing that averaging weights from fine-tuned models can improve accuracy and robustness.

  5. 2025

    PersonalLLM at ICLR

    In 2025, he coauthors PersonalLLM: Tailoring LLMs to Individual Preferences, an ICLR paper on adapting language models to individual preferences.

  6. 2026

    MBABench for LLM agents

    In 2026, he coauthors MBABench, a NeurIPS 2026 paper that evaluates LLM agents on end-to-end finance spreadsheet tasks. The benchmark finds that even the strongest agents fall short of professional finance standards, and performance drops as tasks become harder.

  7. As of October 2026

    Associate professor at Columbia

    As of October 2026, he is an associate professor in Columbia Business School's Decision, Risk, and Operations division and a member of the Columbia Data Science Institute. Before joining Columbia, he spends a year as a research scientist on Meta's Adaptive Experimentation team. His research builds trustworthy AI systems for real-world decision-making.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

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 Hongseok Namkoong yet. Send us a free-licensed photo.

  1. Papers | Hongseok Namkoong hsnamkoong.github.io. Checked 19 October 2026.
  2. Variance-based Regularization with Convex Objectives proceedings.neurips.cc. Checked 19 October 2026.
  3. Fairness Without Demographics in Repeated Loss Minimization proceedings.mlr.press. Checked 19 October 2026.
  4. ORIE Colloquium: Hongseok Namkoong (Stanford) - Learning Models with Uniform Subpopulation Performance events.cornell.edu. Checked 19 October 2026.
  5. Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time proceedings.mlr.press. Checked 19 October 2026.
  6. MBABench: Evaluating LLM Agents on End-to-End Spreadsheet Tasks in Finance arxiv.org. Checked 19 October 2026.
  7. Formal Bio | Hongseok Namkoong hsnamkoong.github.io. Checked 19 October 2026.
  8. Hongseok Namkoong hsnamkoong.github.io. Checked 19 October 2026.