Eric Mitchell
Co-lead of OpenAI's Post-training Frontiers team; co-author of Direct Preference Optimization (DPO)
Eric Mitchell co-leads the Post-training Frontiers team at OpenAI with Yann Dubois, a team that post-trains and deploys large models such as o1, o3, and GPT-5-Thinking. Mitchell earned a PhD in computer science at Stanford, advised by Chelsea Finn and Christopher D. Manning, with work on making language models more trustworthy and easier to use. During that time Mitchell co-authored the Direct Preference Optimization (DPO) paper and was first author of DetectGPT, a zero-shot method for detecting machine-generated text.123
Work
Direct Preference Optimization (DPO) (external site: arxiv.org)Published research
Co-author, with Rafael Rafailov, Archit Sharma, Stefano Ermon, Christopher D. Manning, and Chelsea Finn, of the 2023 paper that introduced DPO for training language models directly from preference data.2
DetectGPT (external site: arxiv.org)Published research
First author of the 2023 paper on zero-shot detection of machine-generated text using probability curvature.3
Other documented roles
Roles the sources document but that are not confirmed as current.
- PhD student in Computer Science, advised by Chelsea Finn and Christopher D. Manning, Stanford University1
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.