← back to paper
arxiv: 2604.07098 · 2 revisions
Selective Neuron Amplification in Transformer Language Models