Use mechanistic interpretability to inspect how trained neural networks process information and make decisions.

Soon you can unlock how this was done.

Behind this: the tool used · the method · what actually resulted.