Adding a feature's direction into a model's activations to deliberately push its behavior.
Continue to AI University →