Andy Arditi

Researchers find single direction controls LLM refusal behavior

Researchers have identified a single, one-dimensional subspace within large language models that is responsible for their refusal to respond to harmful instructions. By manipulating this specific direction in the model'…