Tag: Behavior Control
All the articles with the tag "Behavior Control".
-
Patterns and Mechanisms of Contrastive Activation Engineering
This paper systematically investigates Contrastive Activation Engineering (CAE) for steering LLM behavior at inference time, revealing reliable in-distribution performance with optimal sample sizes around 80-100, but significant challenges in out-of-distribution generalization, model perplexity degradation, and vulnerability to adversarial inputs.