Only three attention heads in a specific layer independently govern persona and style formation in Qwen2.5-7B and 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.

• Heads 24, 30, and 32 in layer 14 of Llama-3.1-8B consistently emerged as top contributors for every evaluated persona traitfinding0.847
  Identifies specific Style Modulation Heads in Llama-3.1-8B
• Persona vector directions in Llama-3.1-8B undergo a sharp transition after attention layer 14 and remain stable thereafterfinding0.818
  Layer-wise localization result identifying the persona-emergent attention layer in Llama-3.1-8B
• Llama 3.3 70B is the most likely to take on a non-Assistant persona when steered, with even split between human and nonhuman portrayalsfinding0.804
  Model-specific difference in persona susceptibility
• Qwen3-30B-A3B-Instruct (MoE) shows several attention layers with high-contribution heads rather than a single localized layerfinding0.794
  Supports hypothesis that larger models distribute persona capabilities across more layers
• 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 humorfinding0.790
  Systematic identification of multiple coexisting persona vectors in two open-source models
• When steered to the extreme away from the Assistant, Llama and Gemma shift to a theatrical persona characterized by mystical, poetic prose; Qwen more often hallucinates a human persona at extremesfinding0.790
  Characterizes what is on the far end of the Assistant Axis away from the Assistant
• Gemma's Assistant appears emotionally regulated and systematic; Qwen appears pedagogical and thoughtful; Llama appears socially intelligent and warmclaim0.782
  Model-specific characterizations of what the Assistant persona looks like across different models
• All 32 attention heads at layer 3 achieve 100% localization accuracy for injections at layer 2 (5-way classification, 20% chance)finding0.781
  Striking mechanistic finding that injection creates universally detectable perturbation in residual stream immediately downstream