Contrastive Steering Vector Construction

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

• Steering vector constructed from all 16 contrastive pairs outperforms most single-pair vectors; best 4-pair vector outperforms full 16-pair vectorfinding0.824
  Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.
• steering vectorsconcept0.808
  A method for modifying model behavior by adding perturbation vectors to activations, used here to try to reduce eval awareness.
• Contrastive Activation Steeringmethod0.799
  Core technique: takes mean difference of model activations on contrastive prompts and adds the resulting vector to the residual stream at inference time.
• Contrastive concept vector extractionmethod0.778
  Method for obtaining concept vectors by subtracting activations from two contrasting prompts.
• Composite Steering Vector (v_role-truth)concept0.776
  Steering vector extracted in Experiment 2 capturing latent representation of desired role behavior and honesty semantics
• Steering vectors capture latent dimensions of reflective behavior more faithfully than surface-level embedding similarity.claim0.758
  Supported by the instruction discovery experiments comparing steering vs. embedding baselines.
• Steering vectors discover effective triggers such as 'However' and 'Otherwise', consistent with prior reported reflection datasetsfinding0.755
  Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
• Bidirectional Steeringconcept0.749
  Ability to steer model behavior in two opposite semantic directions on a trait.