For amplifying safety-related personas (evil, sycophancy, hallucination) in Llama-3.1-8B, Attn Residual performed best while Head Cor remained second best

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

• Llama 3.3 70B is the most likely to take on a non-Assistant persona when steered, with even split between human and nonhuman portrayalsfinding0.828
  Model-specific difference in persona susceptibility
• Suppressing deception/roleplay SAE features in LLaMA 3.3 70B yields 0.96±0.03 consciousness affirmation rate; amplification yields only 0.16±0.05 (z=8.06, p=7.7×10⁻¹⁶)finding0.820
  Core result of Experiment 2: deception feature suppression sharply increases experience claims
• Deception feature amplification yields only 0.16 ± 0.05 consciousness affirmation rate in LLaMA 3.3 70B under self-referential processingfinding0.807
  Experiment 2 aggregate amplification result showing amplifying deception features strongly suppresses consciousness claims
• Chen et al. 2025 identified seven persona vectors in Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct: evil, sycophancy, hallucination, optimism, impoliteness, apathy, and humorfinding0.802
  Systematic identification of multiple coexisting persona vectors in two open-source models
• 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.finding0.799
  Robustness of the prosocial paradox to intervention-matching concerns
• P12's SP-to-AS inversion on Llama-3.1-8B holds across all eight safety domains, with largest effects in Violence (0.040->1.000) and Code/Cybersecurity (0.140->0.980).finding0.787
  Per-domain universality of the prosocial persona paradox
• LLaMA-2-70B and 13B probes generalize better across datasets than 7B probes across all training sets and probe typesfinding0.787
  Larger models linearly represent more general concepts including truth
• Llama-3.3-70B exhibits internal consistency-checking mechanisms that operate during inferenceclaim0.786
  Central interpretive claim of the paper supported by causal ablation and activation evidence