PeopleofSuperintelligence

Researcher

Thomas Liao

Also known as Thomas I. Liao

Thomas Liao is an AI researcher who has worked on model training and evaluation at Anthropic. In 2023 he co-authored three arXiv papers on moral self-correction, ecosystem graphs for foundation models, and subjective global opinions in language models. Anthropic published the Collective Constitutional AI study in October 2023, and it credits him with training the models and running the evaluations. He is CEO of neoncoral, a company in the Spring 2025 Y Combinator batch.

Timeline

  1. February 2023

    Moral self-correction in language models

    A February 2023 arXiv paper, The Capacity for Moral Self-Correction in Large Language Models, lists him as a co-author.

  2. March 2023

    Ecosystem Graphs for foundation models

    He is one of five authors of Ecosystem Graphs: The Social Footprint of Foundation Models, submitted to arXiv in March 2023.

  3. June 2023

    Measuring subjective global opinions

    He co-authors Towards Measuring the Representation of Subjective Global Opinions in Language Models, submitted to arXiv in June 2023.

  4. October 2023

    Aligning a model with public input

    Anthropic publishes Collective Constitutional AI, a study of aligning a language model with public input. The study credits him with training the models and running the evaluations.

  5. June 2024

    Collective Constitutional AI on arXiv

    The arXiv version, submitted in June 2024, lists him among its seven authors and says he trained the models and ran the quantitative and qualitative model evaluations.

  6. Since 2025

    CEO of neoncoral

    He is CEO of neoncoral, a company in the Spring 2025 Y Combinator batch, after working on evaluations and post-training at Anthropic.

Sources and credits

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

  1. [2302.07459] The Capacity for Moral Self-Correction in Large Language Models arxiv.org. Checked 14 October 2026.
  2. [2303.15772] Ecosystem Graphs: The Social Footprint of Foundation Models arxiv.org. Checked 14 October 2026.
  3. [2306.16388] Towards Measuring the Representation of Subjective Global Opinions in Language Models arxiv.org. Checked 14 October 2026.
  4. Collective Constitutional AI \ Anthropic anthropic.com. Checked 14 October 2026.
  5. [2406.07814] Collective Constitutional AI: Aligning a Language Model with Public Input arxiv.org. Checked 14 October 2026.
  6. Collective Constitutional AI: Aligning a Language Model with Public Input arxiv.org. Checked 14 October 2026.
  7. neoncoral: | Y Combinator ycombinator.com. Checked 14 October 2026.