Interpretability for Discovery: Understanding and Discovering Novel Knowledge in AI Models

Researchers investigate whether interpreting AI models can reveal new knowledge. The programme considers methods beyond language-model settings, differences between architectures and data types, and the experiments needed to connect learned representations to useful scientific insights. Session: 08:00–17:00 (America/New_York). A paid NeurIPS virtual pass includes workshop livestreams; on-site access follows the host-location registration rules.