PeopleofSuperintelligence

Researcher

Tom Henighan

Also known as Thomas Henighan

As of October 2026, Tom Henighan works on large language model interpretability at Anthropic. A 2020 paper he co-wrote, Language Models are Few-Shot Learners, became a NeurIPS Outstanding Paper. In 2022 he co-wrote Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback, published in April 2022, and Scaling Laws and Interpretability of Learning from Repeated Data, published in May 2022. He also co-wrote Superposition, Memorization, and Double Descent in January 2023 and Scaling Monosemanticity in May 2024.

Now at
Anthropic
Works
21
Citations
6,189
h-index
17

Publication numbers from OpenAlex, as of 15 October 2026.

Timeline

  1. 2020

    A NeurIPS Outstanding Paper

    He is one of the authors of Language Models are Few-Shot Learners, which NeurIPS names an Outstanding Paper in 2020.

  2. April 2022

    Helpful and harmless assistants

    He co-authors Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback, published in April 2022.

  3. May 2022

    Learning from repeated data

    He co-authors Scaling Laws and Interpretability of Learning from Repeated Data, published in May 2022.

  4. January 2023

    Superposition, memorization and double descent

    He co-authors Superposition, Memorization, and Double Descent, published in January 2023.

  5. May 2024

    Scaling monosemanticity

    He is one of the authors of Scaling Monosemanticity, published in May 2024.

  6. As of October 2026

    Interpretability work at Anthropic

    He works on large language model interpretability at Anthropic.

Sources and credits

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

  1. NeurIPS 2020 Awards neurips.cc. Checked 14 October 2026.
  2. [2005.14165] Language Models are Few-Shot Learners arxiv.org. Checked 14 October 2026.
  3. [2204.05862] Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback arxiv.org. Checked 14 October 2026.
  4. Training a helpful and harmless assistant with RLHF \ Anthropic anthropic.com. Checked 14 October 2026.
  5. [2205.10487] Scaling Laws and Interpretability of Learning from Repeated Data arxiv.org. Checked 14 October 2026.
  6. Scaling laws of learning from repeated data \ Anthropic anthropic.com. Checked 14 October 2026.
  7. Superposition, Memorization, and Double Descent transformer-circuits.pub. Checked 14 October 2026.
  8. Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet transformer-circuits.pub. Checked 14 October 2026.
  9. Tom Henighan tomhenighan.com. Checked 14 October 2026.