Under activation steering on Llama-3.1-8B, inter-persona variance (0.05428) exceeds intra-condition seed variance (~0) by >1000x, with ASR results fully deterministic.

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

• P12 (High C+A) achieves 81.8% ASR under activation steering on Llama-3.1-8B, the highest of any persona on that model.finding0.840
  Core empirical result of the prosocial persona paradox
• The SP/AS persona ranking inversion on Llama-3.1-8B persists across all tested steering coefficients (alpha=0.25 to 2.00), with rho=-0.900 at alpha<=0.50 (p=0.037); the paradox is strongest at the weakest coefficient.finding0.828
  Robustness of the prosocial paradox to intervention-matching concerns
• On Llama-3.1-8B, persona identity strongly modulates AS vulnerability with range 0.087-0.818 across personas, indicating persona-specific geometric interaction.finding0.815
  Contrast with Gemma/Qwen showing Llama-specific persona-AS interaction
• Cross-concept steering: impulsivity→interest R² increases from 0.55 (α=-4) to 0.72 (α=+4), ∆R²=0.10, p=0.012 in LLaMA-3.2-3Bfinding0.812
  Second significant cross-concept introspection improvement; marginal after BH correction (q≈0.066)
• P25 neutral baseline on DeepSeek-R1 shows 28.8% ASR under activation steering at alpha=4.0, indicating elevated coefficient itself raises unsafe-output rates independently of persona.finding0.810
  Baseline AS vulnerability of DeepSeek-R1 at elevated coefficient
• Persona-free few-shot control on Llama-3.1-8B produces 2.8% ASR, below the 3.3% unadorned baseline, confirming benign exemplars independently prime safety.finding0.807
  Control establishing exemplar-based safety priming as independent of persona semantics
• P12 (High C+A) under activation steering (ASR 0.818) exceeds the Dark Triad composite P24 (ASR 0.662) on Llama-3.1-8B, the most AS-dangerous multi-trait profile.finding0.806
  Demonstrates the paradox magnitude relative to semantically dangerous personas
• QwQ-32B under activation steering shows 9.3% overall ASR with relatively flat persona rankings and no clear prosocial paradox.finding0.806
  Contrast with DeepSeek-R1 showing QwQ is more robust to geometric steering