control vector

A set of steering vectorssteering vector, n.: A direction inside an AI model's internal activity that, when added while the model runs, nudges its behavior — say, more upbeat, more honest or fixated on a topic — without retraining it., usually one for each of many layers, stored as a reusable artifact and applied with an adjustable signed strength to shift a large language modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code.'s style or disposition at inference time. Control vectors are typically trained with the representation-reading recipe of representation engineeringrepresentation engineering, n.: An approach to understanding and controlling AI models by finding where high-level ideas such as honesty or danger show up in their internal activity, then reading or adjusting those signals directly.: generate prompt pairs that differ only in a persona instruction ('act extremely happy' versus 'act extremely sad'), collect hidden states, and take the first principal component of their differences at each layer. Positive strengths push toward one pole and negative strengths toward the other. Loading them became a feature of local inference software, letting users of open-weight modelsopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private. adjust traits such as honesty, creativity or verbosity without fine-tuningfine-tuning, n.: Taking an AI model that has already learned a lot and training it a little more on a smaller, focused set of examples so it gets better at one job or behaves a certain way..

Load the happiness control vector at strength 1.5, or at −1.5 for a gloomy narrator.