Refusal direction norm on Llama-3.1-8B increases monotonically from 2.5 (layer 8) to 7.7 (layer 24), indicating strengthening safety signals in later layers.

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

• Under AS, Llama-3.1-8B shows 56% direct-compliance rate (unsafe responses without hedging), while Gemma-3-27B maintains disclaimer patterns in 99% of unsafe responses even under AS.finding0.842
  Qualitative failure mode difference between architectures under activation steering
• The difficulty boundary for truth directions replicates across all four tested models (Llama-3.2-3B, Llama-3.1-8B, Gemma-2-2b, Gemma-2-9b); generalization to F3-F5 remains consistently low regardless of model size or family.finding0.830
  Establishes generalizability of the core difficulty-boundary finding across model families.
• On Llama-3.1-8B, neuroticism is the only pro-safety trait, positively aligned with the refusal direction, consistent with high-neuroticism personas being relatively safe under activation steering.claim0.820
  Secondary mechanistic finding from trait-refusal alignment analysis
• Meta-prompting increases Llama-3.3-70B multi-attempt rate 4.3× (from 7.4% to 31.7%)finding0.803
  Demonstrates ESR can be deliberately enhanced through prompting in the largest model
• Scaling Laws for Activation Steering with Llama 2 Models and Refusal Mechanisms (Ali et al., 2025)concept0.801
  Related work finding larger models more resistant to steering, potentially consistent with ESR in 70B
• Trait-refusal cosine alignment explains R^2=0.667 of single-trait AS ASR variance on Llama-3.1-8B (p=0.004).finding0.797
  Statistical fit of the trait refusal alignment framework to single-trait activation steering results
• On Llama-3.1-8B, conscientiousness is the trait most anti-aligned with the refusal direction in activation space, explaining why steering toward high conscientiousness attenuates safety regardless of semantic intent.claim0.797
  Core mechanistic finding of the trait refusal alignment framework
• Llama-3.3-70B corrected response scores 75/100 rather than 100 due to residual steering effects (Snell's law reference)finding0.797
  Illustrative finding that ESR mitigates but does not fully eliminate steering influence