method
active
method:difference-in-means-steering-vectorsDifference-in-means steering vectors
Used to extract trait and style directions.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (3)
method
- Difference-in-Means Steeringrelated_toSteering vector from mean activation difference between contrasting datasets.
- Causal intervention technique: edit NLA explanation, reconstruct via AR, use difference as steering vector to manipulate model behavior.
- Poison 5% of SFT/DPO data with a trait, then test interventions.
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.
- A method for modifying model behavior by adding perturbation vectors to activations, used here to try to reduce eval awareness.
- Priority claim establishing novelty of the Head Cor intervention approach
- Contrast with probe steering.
- Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.
- Exploratory experiment computing vectors over multiple shots.
- Steering vector extracted in Experiment 2 capturing latent representation of desired role behavior and honesty semantics
- Theoretical constraint proven by Venkatesh and Kurapath (2026): successful steering does not imply the vector is the unique semantic representation
- Paradigm of finding the right direction in activation space (e.g., linear steering).