Samuel R. Bowman
Technical AI safety researcher at Anthropic; NYU associate professor on leave; co-author of Petri
- Technical AI safety researcher, Anthropic1
- Associate Professor of Data Science and Computer Science (on long-term leave); led the Alignment Research Group 2022–24, New York University1
Samuel R. Bowman works on technical AI safety at Anthropic and is on long-term leave from New York University, where Bowman is an associate professor of data science and computer science. Bowman is first author of the SNLI natural language inference corpus and a co-author of the GLUE and SuperGLUE benchmarks, and is the final listed author of Anthropic's Petri alignment auditing tool releases.123467
Career
Bowman earned a PhD in 2016 at the Stanford NLP Group and Stanford Linguistics, supervised by Chris Potts and Chris Manning, for work on early neural network models for text. At New York University, Bowman became an associate professor of data science and computer science, was part of the ML² Group and the CILVR Lab, and led the Alignment Research Group during 2022–24.1
Bowman now works on technical AI safety at Anthropic and is on long-term leave from NYU.1
Research
Bowman's early work includes the SNLI corpus for natural language inference (2015). With Alex Wang and colleagues, Bowman co-authored the GLUE (2018) and SuperGLUE (2019) benchmarks for evaluating general-purpose language understanding. In 2022 Bowman was first author of "Measuring Progress on Scalable Oversight for Large Language Models".2345
Bowman is the final listed author of the October 2025 Petri release, an open-source framework that uses AI agents to audit model behavior, and of the January 2026 Petri 2.0 update.67
Work
Petri (Inspect Petri)Open-source code
Listed as the final author of the October 2025 Petri release and the January 2026 Petri 2.0 update on Anthropic's Alignment Science Blog.67
SNLI corpus (external site: arxiv.org)Open data
First author of "A large annotated corpus for learning natural language inference" (2015), with Gabor Angeli, Christopher Potts and Christopher D. Manning.2
GLUE and SuperGLUE benchmarks (external site: arxiv.org)Published research
Co-author of the GLUE (2018) and SuperGLUE (2019) benchmarks for general-purpose language understanding.34
Career timeline
Aug 2015
Posted the SNLI natural language inference corpus paper as first author.2
2016
Completed a PhD at Stanford, supervised by Chris Potts and Chris Manning.1
Apr 2018
Co-authored the GLUE benchmark for natural language understanding.3
May 2019
Co-authored the SuperGLUE benchmark.4
Nov 2022
Posted "Measuring Progress on Scalable Oversight for Large Language Models" as first author.5
Oct 6, 2025
Co-authored the release of Petri, an open-source alignment auditing tool.6
Jan 22, 2026
Co-authored the Petri 2.0 update with Kai Fronsdal and Jonathan Michala.7
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Amanda AskellAlso at Anthropic
- Aylin CaliskanSafety and alignment
- Abigail SeeAlso from Stanford
- Ami VoraAlso at Anthropic
- Collin BurnsAlso at Anthropic
- Abubakar AbidAlso from Stanford
- Arthur ConmyAlso at Anthropic
- Diogo AlmeidaLanguage models
- Adina WilliamsAlso from NYU
- Awni HannunAlso at Anthropic
- Eric Michael SmithSafety and alignment
- Albert GuAlso from Stanford
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/samuel-r-bowman/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/samuel-r-bowman/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Samuel R. Bowman — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.