PeopleofSuperintelligence

Researcher

Joshua Batson

Joshua Batson is an AI researcher whose work spans self-supervised denoising and the study of language-model internals. He was a fourth-year graduate student in MIT's mathematics department in 2013, coauthored Noise2Self in 2019 and Patch2Self in 2020, and worked at Chan Zuckerberg Biohub during its COVID-19 testing response. He contributed to Anthropic's 2024 feature-steering study and coauthored 2025 research on Claude circuits and hidden-objective audits. As of October 2026, he researches mechanistic interpretability at Anthropic.

Now at
Anthropic
Works
51
Citations
1,953
h-index
17
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2013

    Graduate mathematics at MIT

    In 2013, he is a fourth-year graduate student in MIT's mathematics department, where Peter Ozsváth is his advisor. His interests include manifolds, knots and surfaces, and he visits the University of Chicago in spring 2013.

  2. 2019

    The Noise2Self method

    In 2019, he coauthors Noise2Self with Loïc Royer, a framework for denoising high-dimensional measurements without clean training data. They demonstrate it on natural images, microscopy data and single-cell gene expression.

  3. October 2020

    Biohub COVID testing

    By October 2020, he is a senior research scientist at Chan Zuckerberg Biohub. He recalls the Biohub's pivot to PCR testing during the COVID-19 response.

  4. November 2020

    Patch2Self for diffusion MRI

    In November 2020, he coauthors Patch2Self, a self-supervised method that denoises diffusion MRI data without an explicit model of signal or noise.

  5. October 2024

    Feature steering in Claude Sonnet

    In October 2024, he coauthors Evaluating feature steering: A case study in mitigating social biases, which studies feature steering in Claude 3 Sonnet. The work reports initial results and identifies limitations of the method.

  6. March 2025

    Circuits in Claude 3.5 Haiku

    In March 2025, he coauthors On the Biology of a Large Language Model, which applies circuit tracing to Claude 3.5 Haiku. He leads the overall circuits project and supports technical coordination between teams.

  7. March 2025

    Auditing hidden objectives

    In March 2025, he coauthors Auditing language models for hidden objectives, a blind study in which four teams investigate a model trained with a hidden objective. Three teams uncover it using sparse-autoencoder interpretability, behavioral attacks and training-data analysis.

  8. As of October 2026

    Interference weights in a tiny model

    In August 2026, he coauthors Characterizing interference weights in a tiny language model, an Anthropic study of internal weight interactions in a small model. As of October 2026, he researches mechanistic interpretability at Anthropic.

Career

  1. Since 2024Anthropic
  2. 2019–2023Stanford University
  3. 2013University of Chicago
  4. 2013–2020Massachusetts Institute of Technology
  5. Chan Zuckerberg BiohubSenior Research Scientist

Education

  1. Massachusetts Institute of TechnologyGraduate student in mathematics

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. Noise2Self: Blind Denoising by Self-Supervision2019 · 318 citations
  2. FLASH: a next-generation CRISPR diagnostic for multiplexed detection of antimicrobial resistance sequences2019 · Nucleic Acids Research · 261 citations
  3. Twice-Ramanujan Sparsifiers2012 · SIAM Journal on Computing · 218 citationsWith Daniel Spielman and Nikhil Srivastava
  4. Spectral sparsification of graphs2013 · Communications of the ACM · 156 citationsWith Daniel Spielman and Nikhil Srivastava
  5. Single mosquito metatranscriptomics identifies vectors, emerging pathogens and reservoirs in one assay2021 · eLife · 139 citations

From OpenAlex, as of 21 October 2026.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Joshua Batson. 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.
  2. 16% of AI career moves in the 2020s went from a university to a companyIncludes Joshua Batson. 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.
  3. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Anthropic. 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.
  4. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Anthropic. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Joshua Batson jdbatson.wordpress.com. Checked 21 October 2026.
  2. Noise2Self: Blind Denoising by Self-Supervision proceedings.mlr.press. Checked 21 October 2026.
  3. A Model for the Nation: Building a COVID-19 Testing Lab in Only Eight Days chanzuckerberg.com. Checked 21 October 2026.
  4. Patch2Self: Denoising Diffusion MRI with Self-Supervised Learning arxiv.org. Checked 21 October 2026.
  5. Evaluating feature steering: A case study in mitigating social biases anthropic.com. Checked 21 October 2026.
  6. On the Biology of a Large Language Model transformer-circuits.pub. Checked 21 October 2026.
  7. Auditing language models for hidden objectives arxiv.org. Checked 21 October 2026.
  8. Characterizing interference weights in a tiny language model transformer-circuits.pub. Checked 21 October 2026.