The 'where to steer' dimension is orthogonal and complementary to existing 'how to steer' and 'when to steer' approaches

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

• Where to Steerconcept0.826
  The orthogonal question this paper addresses: which component to intervene on, distinct from how and when to steer
• direction-based steeringconcept0.784
  Paradigm of finding the right direction in activation space (e.g., linear steering).
• steering (intervention on internals)concept0.769
  General technique of modifying activations to control model behavior.
• The core problem of steering should be recast from finding the right direction to finding the right geometryclaim0.765
  The paper's programmatic conclusion about how the field should reconceptualize neural network steering
• This is the first study to mitigate the performance trade-off in steering vectors derived from the difference-in-means approachclaim0.759
  Priority claim establishing novelty of the Head Cor intervention approach
• 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 themselvesclaim0.754
  Central interpretive claim organizing the entire paper's results
• steering vectorsconcept0.752
  A method for modifying model behavior by adding perturbation vectors to activations, used here to try to reduce eval awareness.
• Steering vectors from µ(0→2) slightly outperform µ(1→2) for instruction discovery across datasets and modelsfinding0.752
  Shows that contrasting No Reflection with Triggered Reflection provides a stronger signal than Intrinsic vs Triggered.