A camp in Model Interpretability that states we should only focus on specific sections of a model rather than the whole model.

  • Learn which parts are responsible for halucinations