Research

Interpretability and explainability

How models work inside, and how to tell whether an explanation of them is true. We look inside language models, and ask when we can believe what a model says about itself, or what our tools say about it. The questions are often as much philosophical as technical.

Mechanistic interpretability · Philosophy of machine learning · Explanation · Evaluation

6 papers, people →

Structured and scientific data

Machine learning for data that is not text or images: tables, knowledge graphs and measurements from the environment. We work on finding the right dataset, describing it honestly, and forecasting from it.

Dataset search · Knowledge graphs · Retrieval · Environmental data

4 papers, people →