Dose-Response Feature Steering Protocol

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Interpretability-Driven Feedback Steeringconcept0.726
  Framework of using internal-state representations to control or steer generative models; conceptually parallel to manifold steering in language models.
• Our method enables bidirectional steering of model behavior.finding0.708
  The method can steer the model in both positive and negative directions on the target semantic.
• Feature steering (clamping feature activations)method0.704
  Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
• steering (intervention on internals)concept0.700
  General technique of modifying activations to control model behavior.
• How can we be sure that steering methods actually elicited the deployment behavior, as opposed to only suppressing verbalizations of being deployed?question0.696
  Central motivating question of the paper; the model organism approach is the proposed answer.
• Contrastive Activation Steeringmethod0.695
  Core technique: takes mean difference of model activations on contrastive prompts and adds the resulting vector to the residual stream at inference time.
• Steering vector extracted from final post-expert-iteration model also successfully elicits deployment behaviorfinding0.695
  Replicates main result using in-distribution steering vector; addresses concern about pre-trained vector validity.
• Steering's effect on verbalized chain-of-thought evaluation/deployment beliefs is a useful proxy for its effect on actual deployment behaviorclaim0.695
  Practical guidance for practitioners who lack ground-truth model organisms.