Patrick Rebeschini
Rebeschini earned a Princeton doctorate in operations research and financial engineering and held Yale research appointments before Oxford. His research combines probability, statistics and optimization. His work includes model aggregation, learned feature geometry and contextual bandits. These study how learning methods use data and generalize, with mathematical assumptions stated in the papers. He coauthored Knowing When to Quit, an ICML 2026 study on dynamic abstention in language-model reasoning. This connects decision rules with reliable model behavior.
- Now at
- University of Oxford
- Works
- 66
- Citations
- 419
- h-index
- 8
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
As of October 2026
Training and research career
Rebeschini earned a Princeton doctorate in operations research and financial engineering and held Yale research appointments before Oxford. His research combines probability, statistics and optimization.
-
As of October 2026
Learning and generalization
His work includes model aggregation, learned feature geometry and contextual bandits. These study how learning methods use data and generalize, with mathematical assumptions stated in the papers.
-
2026
When language models should abstain
He coauthored Knowing When to Quit, an ICML 2026 study on dynamic abstention in language-model reasoning. This connects decision rules with reliable model behavior.
Sources and credits
Photo: ellis.eu; Patrick Rebeschini profile photograph; 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.