Edward Grefenstette
Edward Grefenstette researches natural language, reasoning and learning. His Oxford DPhil, examined in 2013, studied compositional models of language meaning. He co-founded Dark Blue Labs in 2014 and then held research posts at DeepMind. With Richard Evans, he coauthored Learning Explanatory Rules from Noisy Data, connecting differentiable learning with logical rules. He later worked at Facebook and Cohere and became a Google DeepMind research director in June 2023. His interests include reinforcement learning, meta-learning and reward modeling.
- Now at
- Google DeepMind
- Works
- 109
- Citations
- 12,263
- h-index
- 31
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2013
Language meaning in doctoral work
Grefenstette completed an Oxford DPhil in June 2013. His thesis studied compositional distributional models of natural-language meaning. This links mathematical structure with ways of representing the meaning of words and sentences.
-
2014–2018
Moving into AI laboratories
He co-founded Dark Blue Labs and served as its CTO in 2014, then began research positions at DeepMind in November 2014. These periods overlap with his Oxford fellowship.
-
2017–2018
Learning explanatory rules
With Richard Evans, Grefenstette proposed a differentiable inductive-logic framework. The work addresses learning logical rules in the presence of noisy data and connecting the system with neural networks. The preprint appeared in 2017 and the work appeared in JAIR in 2018.
-
2018–2023
Research across organizations
He worked in research at Facebook from December 2018 to March 2022 and led machine learning at Cohere from April 2022 to June 2023. He became a Google DeepMind research director in June 2023. His research interests include reinforcement learning, meta-learning and reward modeling.
Sources and credits
Photo: ELLIS profile; creator not stated. License: Not stated by source. Source: ellis.eu. Cropped to a square and resized. Ask us to remove or replace this photo.