What a model exposes by default tracks the norms it was trained toward, while steering acts on deviations from those defaults rather than on the defaults themselves

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

• Models are not merely tracking dialogue context features; same-concept steering shows privileged internal access is necessary to explain self-report shiftsclaim0.837
  Addresses skeptical alternative that reports reflect only conversational content
• Which behaviors does a model represent internally, default to, can be pushed to amplify, or refuses to expose?question0.827
  The motivating diagnostic question that prompting alone cannot answer
• Optimally steering model behavior requires isolating concept geometry and defining operators to navigate it.claim0.811
• Per-model activation-steering safety verification is necessary because prompt-side persona rankings do not predict geometric vulnerability.claim0.804
  Practical implication drawn from the prosocial persona paradox finding
• Steering base models along the assistant axis extracted from instruct models promotes helpful human archetypes and agreeable traits, suggesting the axis predates post-trainingfinding0.804
  Validation finding from Lu et al. 2026 supporting PSM's claim about pre-training persona structure
• Steering preferentially amplifies exaggerated and undesirable styles, not competence-oriented or assistant-default behaviorsclaim0.801
  The top of the steerability ranking is dominated by exaggerated or attention-grabbing styles
• Steering affects type hint writing by modifying the model's belief about whether it is being evaluated, not by directly encoding type hint informationclaim0.798
  Mechanism claim supported by transcript analysis and the fact that the steering vector was extracted from a model that never writes type hints.
• Activation steering effectively biases latent representations but does not fully replicate the mechanisms triggered by explicit instruction.claim0.796
  Nuanced interpretive claim about the limits of steering as a mechanism for reflection enhancement.