Activation steering is not equivalent to simply adding the deployment-side contrastive prompts to the user message

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

• Contrastive Activation Steeringmethod0.839
  Core technique: takes mean difference of model activations on contrastive prompts and adds the resulting vector to the residual stream at inference time.
• For future powerful models, situational awareness might 'corrupt' the deployment representation in contrastive prompts, potentially rendering contrastive steering ineffectivehypothesis0.820
  Future threat to the method: a highly sophisticated model might be suspicious of deployment-framed prompts during extraction.
• Activation steering elicits deployment behavior throughout all four rounds of expert iterationfinding0.816
  Shows steering remains effective even as model becomes more robust to prompting-based attempts to elicit deployment behavior.
• Activation steering can make an evaluation-aware model act as if deployed, not merely suppress verbalizations of evaluation awarenessclaim0.815
  Central claim of the paper; supported by the model organism ground-truth approach.
• How can we be sure that steering methods actually elicited the deployment behavior, as opposed to only suppressing verbalizations of being deployed?question0.811
  Central motivating question of the paper; the model organism approach is the proposed answer.
• Can activation steering elicit deployment behavior from a model that only believes it is deployed when given a specific deployment cue, even when prompting alone cannot?question0.810
  Future work direction: the inverse problem to the Wood Labs evaluation cue tested in this paper.
• Activation steering effectively biases latent representations but does not fully replicate the mechanisms triggered by explicit instruction.claim0.805
  Nuanced interpretive claim about the limits of steering as a mechanism for reflection enhancement.
• Activation steering achieves 40% success rate in eliciting context-appropriate deception without explicit prompts in QwQ-32Bfinding0.804
  Key intervention result showing steering vectors can induce deceptive behavior from a neutral baseline