concept
active
concept:persona-formation

Persona Formation

The process by which LLMs develop stylistic and behavioral characteristics, which this paper localizes to specific attention heads

Neighborhood — ranked by edge-count

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

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

  • The process of building a coherent model persona from character archetypes and traits during training
  • Personaconcept0.794
    Stable, reidentifiable dispositional profiles that can be interpreted as characters with broadly coherent beliefs, values and traits
  • Persona Fidelityconcept0.773
    The degree to which an LLM's generated responses consistently reflect an assigned persona
  • Persona driftconcept0.773
    Behavioural drift in multi-turn LLM interaction; documented in prior work for persona, identity, and instruction-following
  • Keeping a model anchored to its intended persona during deployment, preventing drift to harmful behaviors
  • Framework by Marks et al. proposing that models infer a context-appropriate persona for next-token prediction and post-training concentrates distribution around helpful assistant
  • Key open question identified for future work on pretraining data as mechanism