Stephen Roller
Member of Technical Staff at Thinking Machines Lab; co-first author of OPT; dialogue and LLM work at Meta FAIR
- Member of the Technical Staff, Thinking Machines Lab1
Stephen Roller's homepage lists a current role as a Member of the Technical Staff at Thinky, linking to the Thinking Machines Lab website. Roller was previously a Senior Staff Research Scientist at Google DeepMind, a Member of the Technical Staff at Character.AI working on pretraining and training infrastructure, and, from 2017 to 2023, a Senior Staff Research Engineer at Meta's Fundamental AI Research (FAIR) team, where Roller worked on the large language model and dialogue teams. At FAIR, Roller was one of three equal-contribution first authors of OPT (2022) and first author of Recipes for Building an Open-Domain Chatbot (2020), which made its models and code publicly available; Roller also co-authored BlenderBot 3 (2022).1245
Work
OPT and metaseq (external site: github.com)Open-source code
One of three equal-contribution first authors of OPT: Open Pre-trained Transformer Language Models (2022), with Susan Zhang and Naman Goyal; the paper releases code for experimenting with the models, and the metaseq repository is published under the MIT License.231
Open-domain chatbot research (Recipes and BlenderBot 3) (external site: arxiv.org)Published research
First author of Recipes for Building an Open-Domain Chatbot (2020), which builds 90M, 2.7B, and 9.4B parameter models and makes the models and code publicly available, and co-author of BlenderBot 3 (2022), a 175B parameter dialogue model whose paper says the weights and code were released.451
Other documented roles
Roles the sources document but that are not confirmed as current.
- Senior Staff Research Scientist, Google DeepMind1
- Member of the Technical Staff (pretraining and training infrastructure), Character.AI1
- Senior Staff Research Engineer (2017–2023), Meta (Fundamental AI Research, FAIR)12
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.