Steering vectors can work as activation oracles! Inspired by @isolyth.dev 's recent work on instruct vectors, I've ported the same approach to create Activation Oracles. Seems to work surprisingly well given the param deficit, nearly matching Activation Oracle LoRA performance.
Post
tl;dr: collect a model's activations, inject those into the activation oracle and ask it natural language questions, get answers. works across finetunes/LoRA variants. Anthropic's release of the original technique has some good diagrams: alignment.anthropic.com/2025/activat...
alignment.anthropic.comActivation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers