Glossary · Safety

Interpretability

The research area focused on understanding what's actually happening inside a model's internal computations, distinct from explainability's focus on justifying outputs after the fact.