Yuan Yuan 袁源

Researcher at OpenAI

Yuan Yuan

AI for scientific discovery.

I’m a researcher on the Rosalind team at OpenAI. My current focus is AI for science—how AI can accelerate discovery in biology, medicine, and empirical science.

Contact

Google Scholar X / Twitter LinkedIn

My former UC Davis email is no longer active. Please reach out on X or LinkedIn.

Portrait of Yuan Yuan
Yuan Yuan / 袁源

Background

Before OpenAI, I was a tenure-track assistant professor at UC Davis (2024–2025) and Purdue University (2021–2024). I received my PhD from MIT’s Institute for Data, Systems, and Society, and bachelor’s degrees in Computer Science and Economics from Tsinghua University.

My previous industry research experience includes a visiting research role at Microsoft and a research internship at Facebook Core Data Science.

Research

What I work on

Current · Rosalind

Accelerating empirical science

My work focuses on AI for scientific discovery.

About Rosalind

Previously · Safety Training

From refusals to safe completions

Previously, I worked on OpenAI’s Safety Training team, contributing to a shift from refusal-based training toward safe completions: making model responses helpful while respecting safety constraints.

Read about safe completions

Earlier research

Networks, experimentation, and social behavior

Previously, I worked as an interdisciplinary researcher at the intersection of AI and empirical science, publishing in general science journals, business-school journals, and computer science conferences.

My earlier work explored social and organizational networks, causal inference and online experiments, and the use of large language models in social science.

Publications

Google Scholar

AI Tech Reports

3

General science journals

4

Business-school journals

4

Computer science

6

Working papers

2

Preprints and manuscripts from my earlier research. Conference and journal versions are linked together above.

* Equal contribution. Conference versions are identified separately from their journal extensions.