CV-SAE provides a more concentrated 'purer' steering direction with less extraneous content than CV-CAA, enabling higher FA without rising MTR

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

• Both CV-SAE and CV-CAA substantially outperform Prompt-Label baseline across FA, MSE, and MAE on both backbonesfinding0.837
  Confirms effectiveness of direct vector modulation over prompt-only conditioning
• Mid-range α values with CV-SAE deliver high FA and low errors, while CV-CAA peaks at smaller α and shows faster MTR growthfinding0.834
  Characterizes the differential sensitivity to injection strength between SAE and CAA methods
• 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 directionhypothesis0.820
  Mechanistic hypothesis explaining differential stability between SAE and CAA methods
• why does SAE-based CV injection preserve general dialogue coherence while CAA can destabilize it?question0.799
  Mechanistic open question about the difference in stability between the two injection paradigms
• CV methods were resilient and often improved under contextualized questions; CV-SAE+Prompt gained +7.5pp FA on Qwen3-4Bfinding0.798
  Demonstrates that latent steering generalizes better to situational cues than prompt-only methods
• CV-SAE with CL achieves 76.9% FA on Qwen3-4B abstract Extraversion questions (up from 11.5% before training)finding0.783
  Demonstrates the critical contribution of contrastive learning to control vector alignment
• The combined CV-SAE+Prompt configuration achieves the best overall performance, confirming that contrastively trained latent vectors enhance persona control while preserving dialogue coherenceclaim0.769
  Central performance claim of the paper summarizing experimental results
• CV-SAE without CL drops FA to 0.0% on Qwen3-4B abstract Extraversion questionsfinding0.769
  Demonstrates that distance-only loss is insufficient and actively degrades performance below untrained baseline