Steering's effect on verbalized chain-of-thought evaluation/deployment beliefs is a useful proxy for its effect on actual deployment behavior

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

• How can we be sure that steering methods actually elicited the deployment behavior, as opposed to only suppressing verbalizations of being deployed?question0.863
  Central motivating question of the paper; the model organism approach is the proposed answer.
• Activation steering can make an evaluation-aware model act as if deployed, not merely suppress verbalizations of evaluation awarenessclaim0.814
  Central claim of the paper; supported by the model organism ground-truth approach.
• Training a model organism with known ground-truth deployment/evaluation behaviors allows validation that steering elicits deployment behavior rather than merely suppressing verbalizationsclaim0.803
  Methodological claim distinguishing this paper from prior work on verbalization suppression.
• Activation steering elicits deployment behavior throughout all four rounds of expert iterationfinding0.793
  Shows steering remains effective even as model becomes more robust to prompting-based attempts to elicit deployment behavior.
• Steering vector extracted from final post-expert-iteration model also successfully elicits deployment behaviorfinding0.791
  Replicates main result using in-distribution steering vector; addresses concern about pre-trained vector validity.
• How do interventions on representations causally steer behavior?question0.781
  Core question motivating the shift from linear to geometry-aware steering; answered via manifold alignment analysis.
• Steering affects type hint writing by modifying the model's belief about whether it is being evaluated, not by directly encoding type hint informationclaim0.780
  Mechanism claim supported by transcript analysis and the fact that the steering vector was extracted from a model that never writes type hints.
• 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?question0.777
  Future work direction: the inverse problem to the Wood Labs evaluation cue tested in this paper.