Steering's effect on verbalized evaluation/deployment beliefs in chain-of-thought is highly correlated with its effect on type hint rate across hyperparameter configurations

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?question0.823
  Central motivating question of the paper; the model organism approach is the proposed answer.
• Activation steering toward deployment reduces type hint rate to <11% even with Wood Labs evaluation cue presentfinding0.810
  Main result: steering elicits deployment behavior even when the evaluation cue is present and prompting fails.
• Activation steering can make an evaluation-aware model act as if deployed, not merely suppress verbalizations of evaluation awarenessclaim0.787
  Central claim of the paper; supported by the model organism ground-truth approach.
• Classifier-free guidance does not decrease type hint rate to deployment levels; activation steering is more effectivefinding0.781
  Comparative result showing steering superiority over CFG as alternative intervention.
• Steering at 6 layers (strength 0.6 each, total 3.6) outperforms single-layer steering at equivalent total strength for type hint suppressionfinding0.777
  Demonstrates distributed steering is more effective and less accuracy-damaging than concentrated steering.
• Linear steering produces noisy off-target effects; manifold steering cleanly shifts probability mass between sequential concepts.finding0.775
  Core empirical claim comparing steering approaches on cyclic concepts.
• Steering vector extracted from final post-expert-iteration model also successfully elicits deployment behaviorfinding0.773
  Replicates main result using in-distribution steering vector; addresses concern about pre-trained vector validity.
• Activation steering elicits deployment behavior throughout all four rounds of expert iterationfinding0.769
  Shows steering remains effective even as model becomes more robust to prompting-based attempts to elicit deployment behavior.