LIME: Local Interpretable Explanations

LIME (Local Interpretable Model-agnostic Explanations) explains a single prediction by fitting a simple, interpretable model (like linear regression) to the complex model behavior in a small neighborhood around that one instance.

A black-box model may be wildly nonlinear overall, but near one point it is approximately linear. LIME exploits this: it does not explain the whole model, only its behavior locally around the instance you care about.