Single-Trait Steerability Classification

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

• Steerable trait (latent but amplifiable)concept0.835
  A trait minimally expressed at baseline but amplifiable when steered (alpha >= 0.5)
• steerable emotion featuresconcept0.789
  Emotion-encoding directions in LLM activation space that can be amplified or suppressed via activation steering to causally drive model behavior
• steering vectorsconcept0.762
  A method for modifying model behavior by adding perturbation vectors to activations, used here to try to reduce eval awareness.
• Natural-steerable pairs split 26 constructive / 22 dominant but produce zero destructive outcomes; natural trait acts as anchorfinding0.756
  Natural trait acts as anchor in mixed pairs, suppressing steerable partner at worst but never collapsing both
• Steerability Score (Phi)concept0.755
  Aggregate metric averaging mean SJT scores across OCEAN traits and steering directions; maximum possible is 10
• Steering vector constructed from all 16 contrastive pairs outperforms most single-pair vectors; best 4-pair vector outperforms full 16-pair vectorfinding0.753
  Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.
• All-token steeringmethod0.738
  Baseline steering method that applies intervention at every token generation step, shown to degrade performance at high strengths
• SEAL (Steerable Reasoning Calibration)framework0.737
  Prior work using steering vectors to control reflection, motivated by reducing redundant self-reflection in long CoT.