Interpretability — seeing inside the model
Also called: mechanistic interpretability, opening the black box, what's going on inside
The research effort to understand what actually happens inside a model's billions of parameters — identifying the internal patterns ('features') and pathways it uses to turn an input into an output, instead of treating it as an unexplained black box.
A model can't reliably tell you why it answered as it did; its stated reasons are more generated text. Interpretability aims for evidence from the internals instead — the early beginnings of a way to audit, debug, and catch problems that behaviour alone hides.
Asking the model to explain itself. A self-explanation is another prediction, not a readout of the internal computation; interpretability tries to inspect the mechanism directly.