Modifying a model's weights during training so that a specific later fine-tuning attempt fails to teach it a targeted capability or behavior, while leaving its normal performance intact.
Continue to AI University →