Steering Llama 2 via Contrastive Activation Addition

arXiv:2312.06681v3 Announce Type: replace-cross
Abstract: We introduce Contrastive Activation Addition (CAA), an innovative method for steering language models by modifying their activations during forward passes. CAA computes "steering vectors" by averaging the difference in residual stream activations between pairs of positive and negative examples of a particular behavior, such as factual versus hallucinatory responses. During inference, these steering vectors are added at all token positions after the user's prompt with either a positive or negative coefficient, allowing precise control over …

abstract arxiv behavior cs.ai cs.cl cs.lg difference examples language language models llama llama 2 negative positive residual type vectors via