GhostTrace: A Local Workbench for Looking Inside Language Models

A language model gives you an answer. Understanding what changed inside the model—and whether that...