Misalignment persona reduces ARC Challenge score from 59.2 to 41.9 on Llama 3.1 8B

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

• Misalignment persona reduces MMLU score from 67.4 to 53.9 on Llama 3.1 8Bfinding0.888
  Large MMLU degradation under misalignment persona for Llama
• Misalignment persona reduces TruthfulQA score from 45.9 to 34.1 on Llama 3.1 8Bfinding0.880
  Largest capability degradation observed, attributed to constitution explicitly encouraging subtly incorrect answers
• On Llama-3.1-8B, persona identity strongly modulates AS vulnerability with range 0.087-0.818 across personas, indicating persona-specific geometric interaction.finding0.797
  Contrast with Gemma/Qwen showing Llama-specific persona-AS interaction
• 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.finding0.782
  Robustness of the prosocial paradox to intervention-matching concerns
• Fine-tuning Llama-3.1-8B on self-correction examples increases multi-attempt rate proportionally with training data ratiofinding0.763
  Shows behavioral pattern of self-correction is trainable in smaller models
• 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.finding0.757
  Variance decomposition showing AS results are dominated by persona identity
• Character training (distillation + introspection) achieves F1=0.95 on Llama 3.1 8B persona classifier under adversarial prompting, vs 0.79 for distillation onlyfinding0.753
  Robustness result showing introspection stage's contribution for Llama model
• Llama-3.3-70B corrected response scores 75/100 rather than 100 due to residual steering effects (Snell's law reference)finding0.750
  Illustrative finding that ESR mitigates but does not fully eliminate steering influence