MATS 学员:
Nick Gabrieli, Nina Panickssery (née Rimsky), Julian Schulz, Meg Tong
作者:
Nina Panickssery, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, Alexander Matt Turner
引用次数
摘要:
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 the degree of the targeted behavior. We evaluate CAA's effectiveness on Llama 2 Chat using multiple-choice behavioral question datasets and open-ended generation tasks. We demonstrate that CAA significantly alters model behavior, is effective over and on top of traditional methods like finetuning and system prompt design, and minimally reduces capabilities. Moreover, we gain deeper insights into CAA's mechanisms by employing various activation space interpretation methods. CAA accurately steers model outputs and sheds light on how high-level concepts are represented in Large Language Models (LLMs).
MATS 是一个独立的研究与教育项目,致力于连接新兴研究人员与 AI 对齐、治理和安全领域的导师。
每期 MATS 项目在加州伯克利进行,为期 12 周。部分学员可选择之后前往伦敦参加为期 6 至 12 个月的延伸项目。