A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models Paper • 2605.08513 • Published 19 days ago • 15
