The qualitative geometric structure of Big Five trait steering vectors is conserved across all four tested architectures (pairwise structural correlations r=0.898-0.986), motivating treating multi-trait steering as first-order additive.

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

• Pairwise structural correlations of inter-trait cosine similarity matrices across all four architectures range from r=0.898 to r=0.986 (all p<0.001), indicating cross-architecture geometric preservation.finding0.822
  Cross-architecture geometric invariance of Big Five steering vectors
• Steering vector constructed from all 16 contrastive pairs outperforms most single-pair vectors; best 4-pair vector outperforms full 16-pair vectorfinding0.812
  Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.
• Steering vectors capture latent dimensions of reflective behavior more faithfully than surface-level embedding similarity.claim0.799
  Supported by the instruction discovery experiments comparing steering vs. embedding baselines.
• Neuroticism steering vector has mean cosine alignment of +0.079 with the refusal direction on Llama-3.1-8B, the only pro-safety trait among all five OCEAN dimensions.finding0.798
  Mechanistic finding explaining why high-N personas are safe under steering
• There is a clear bidirectional relationship between the geometry of behavior and representation: steering along representation manifolds follows behavior manifolds, and vice versa.claim0.791
  The paper's finding that the alignment holds in both directions — from representation to behavior and from behavior back to representation space.
• We hypothesize that geometric explanations for persona-safety interactions are tractable and that per-model geometric audits can predict activation-steering vulnerability from trait-refusal cosine alignment.hypothesis0.789
  Forward-looking claim about the utility of the trait refusal alignment framework as a general tool
• Trait space requires 4 dimensions (Gemma, Qwen) and 7 dimensions (Llama) to explain 70% of variance, with distinctive PC1 spanning conscientious to impulsive traitsfinding0.780
  Corroborates role space findings using traits; shows PC1 also captures Assistant-ness in trait space
• What shared semantic or qualitative factor explains why some steering vectors produce more salient and detectable perturbations than others?question0.779
  Open question arising from the 100% accuracy on specific concept-layer-strength combinations