apple-aiml-research / ml-refusal-neurons

git clone https://github.com/apple-aiml-research/ml-refusal-neurons.git

Reproduction code for "A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models"