1 paper
Shivam Adarsh, Hetianyu Huang, Lau Safin +2
Activation steering is an inference-time technique used to control or modify the behavior, tone, or safety properties of a large language model (LLM) without retraining the model's…