As α increases in CV-CAA, off-target semantic components are introduced alongside the intended concept direction, degrading response quality and inflating MTR

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

• We hypothesize that the control signal learned by CV-CAA is less 'pure' than CV-SAE; as α increases, off-target semantic components are introduced alongside the intended concept directionhypothesis0.852
  Mechanistic hypothesis explaining differential stability between SAE and CAA methods
• Mid-range α values with CV-SAE deliver high FA and low errors, while CV-CAA peaks at smaller α and shows faster MTR growthfinding0.819
  Characterizes the differential sensitivity to injection strength between SAE and CAA methods
• what α value provides the optimal balance between steering accuracy and response stability for different CV methods?question0.756
  Practical hyperparameter question motivating the injection strength sweep experiment
• CV-SAE provides a more concentrated 'purer' steering direction with less extraneous content than CV-CAA, enabling higher FA without rising MTRclaim0.753
  Author's interpretation of why SAE outperforms CAA in stability at higher injection strengths
• In Target−α configuration, Head Cor succeeded in suppressing traits to levels unreachable by other methods while maintaining significantly higher coherencyfinding0.751
  Demonstrates unique capability of Head Cor for trait suppression scenarios
• Both CV-SAE and CV-CAA substantially outperform Prompt-Label baseline across FA, MSE, and MAE on both backbonesfinding0.745
  Confirms effectiveness of direct vector modulation over prompt-only conditioning
• Directing response attention to complement syntax and/or mental state verbs (MSV) yields no significant alterations in IIT estimates compared to entire stimulus analysis.finding0.743
  Suggests LLMs do not represent complement/MSV linguistic features in the same way as they are crucial for human ToM development.
• Contrastive Activation Addition (CAA)method0.743
  An existing activation steering method used as comparative baseline.