concept
active
concept:direction-based-steeringdirection-based steering
Paradigm of finding the right direction in activation space (e.g., linear steering).
Neighborhood — ranked by edge-count
Methods (1)
method
- linear steeringimplementsTypical approach that adds a scaled steering vector to representations; the paper argues this is mismatched with actual representation geometry.
Concepts (1)
concept
- geometry-based steeringrelated_toParadigm of finding the right geometry (manifold) for principled control.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Novel method that applies intervention only when the model begins a new thinking step (at the \n\n delimiter) rather than at every token
- Ability to steer model behavior in two opposite semantic directions on a trait.
- A method for modifying model behavior by adding perturbation vectors to activations, used here to try to reduce eval awareness.
- General technique of modifying activations to control model behavior.
- Latent intervention technique that manipulates sparse features to steer model predictions toward desired concepts.
- Using interventions to guide model generation behavior, e.g., adding sentiment vectors at inference time
- At each step, choose the action that most intensifies the feeling of the emerging whole.
- General approach of using interpretability feedback to steer model generation.