Research
Research Programme
We develop methods for applying language models to social science, with a focus on synthetic data, opinion prediction, and model evaluation.
Papers
Our current research focuses on using LLMs as condensed representations of internet-scale text from which we can learn how people organise relationships between concepts. Papers in progress. Details coming soon.
Code & Data
synthetic_sampling
Research code for synthetic sampling methods. LLM-based opinion prediction and benchmarking.
View on GitHub → Materialsoxford-llms-workshop
Summer school materials: lecture slides, notebooks, and code samples from 2023 to 2025.
View on GitHub → CourseIntro-to-LLMs-DPIR
Teaching materials for the Oxford / Paris / Florence courses on LLMs for social science.
View on GitHub →HuggingFace
Models and datasets from the collaborative research projects are published on the Oxford LLMs HuggingFace profile.