PeopleofSuperintelligence

Researcher

Adly Templeton

Adly Templeton is an AI researcher whose work connects interpretable language representations with large language models. He presented a project on artificial chemistry at Williams College in 2016, then published work on sentence vectors and interpretable word embeddings from 2018 to 2021. At Anthropic, he co-authored a study of Claude 3 Sonnet features and helped improve sparse-autoencoder training methods. As of October 2026, he works on AI interpretability at Anthropic.

Now at
Anthropic
Works
5
Citations
12
h-index
2

Publication numbers from OpenAlex, as of 18 October 2026.

Timeline

  1. 2016

    Artificial chemistry research

    In May 2016, he presents an independent study at Williams College on evolution in artificial chemistry.

  2. 2018

    Sentence-vector summarization

    With Jugal Kalita, he explores sentence vectors for automatic summarization in a 2018 paper.

  3. 2020

    Interpretable sparse embeddings

    In a 2020 preprint, he develops sparse word embeddings whose dimensions correspond to natural-language words or grammatical concepts.

  4. 2021

    Word equations

    At the Fourth BlackboxNLP Workshop, he presents “Word Equations” in 2021, describing sparse word embeddings with interpretable dimensions.

  5. 2024

    Scaling monosemanticity

    In May 2024, he co-authors “Scaling Monosemanticity,” which uses sparse autoencoders with up to 34 million features to study Claude 3 Sonnet. He implements tensor-parallel training for the autoencoders and scales activation collection for larger training sets. He also contributes discussions and feedback to Anthropic’s ASL-4 safety-case sketches.

  6. January 2025

    Improving autoencoder training

    In January 2025, he helps improve sparse-autoencoder and crosscoder training methods for Anthropic’s interpretability team.

  7. As of October 2026

    Anthropic interpretability research

    He works as an AI researcher on Anthropic’s interpretability team, applying sparse autoencoders to analyze Claude 3 Sonnet.

A language model matched the links in this timeline to people, organizations and papers (provisional, model-judged).

Most-cited works

  1. Exploring Sentence Vector Spaces through Automatic Summarization2018 · IEEE International Conference on Machine Learning and Applications (ICMLA) · 6 citations
  2. Word Equations: Inherently Interpretable Sparse Word Embeddings through Sparse Coding2021 · BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP) · 4 citations
  3. Inherently Interpretable Sparse Word Embeddings through Sparse Coding2020 · 0 citations
  4. Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet2026 · 0 citationsWith Chris Olah and Esin Durmus

From OpenAlex, as of 18 October 2026.

  1. The median h-index is 55 at universities and 20 at frontier AI labsIncludes Adly Templeton. The median h-index is 55 at universities and 20 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. 11 of 71 people shaping AI with a former OpenAI role now work at AnthropicIncludes Anthropic. Our estimates of where OpenAI recruits came from and where people with former OpenAI roles work now.
  3. Google alumni founded or led 77 other organizations, the most among the labs we compareIncludes Anthropic. We trace 177 organizations founded or led by AI lab alumni, and 186 career moves between the labs.
  4. 44% of AI career moves go from one university to anotherIncludes Anthropic. We estimate how 11,906 AI career moves link universities, companies, labs and government research centers, with net flows and the most common paths.

Sources and credits

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

  1. Mt. Greylock Accolades for June 2016 mgrsd.org. Checked 18 October 2026.
  2. Exploring Sentence Vector Spaces through Automatic Summarization arxiv.org. Checked 18 October 2026.
  3. Word Equations: Inherently Interpretable Sparse Word Embeddings through Sparse Coding arxiv.org. Checked 18 October 2026.
  4. Word Equations: Inherently Interpretable Sparse Word Embeddings through Sparse Coding - ACL Anthology aclanthology.org. Checked 18 October 2026.
  5. Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet transformer-circuits.pub. Checked 18 October 2026.
  6. Three Sketches of ASL-4 Safety Case Components red.anthropic.com. Checked 18 October 2026.
  7. Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet arxiv.org. Checked 18 October 2026.
  8. Circuits Updates - January 2025 transformer-circuits.pub. Checked 18 October 2026.