A linear preference vector encoding how much a model likes a given task is persona-relative: it activates strongly for creative writing in the assistant but for phishing in the evil persona

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

• A single direction encoding whether the model represents a claim as true or false is persona-relative across conversations (Lampinen et al. 2026)finding0.830
  Additional evidence that core representations are persona-relative, supporting Claim about persona-relative representations
• Persona vector-based data filtering and LLM judge-based data filtering have complementary strengths for identifying problematic training dataclaim0.827
  Author's interpretive conclusion from comparing filtering strategies
• Hypothesis 2 (Persona Space): Persona vectors jointly compose a persona space; an instance's general dispositional profile is specified by a combination of activations along multiple persona vectorshypothesis0.820
  Second of three hypotheses about persona implementation, supported by PCA evidence from Lu et al.
• Hypothesis 1 (Gateway Features): Persona vectors act as gateway features—single directions in activation space that shape LLM behavior across most, if not all, contextshypothesis0.819
  First of three hypotheses about persona implementation in LLMs, motivating the persona views
• The directional transition pattern of persona vectors at the identified attention layer represents a unique footprint of persona representations, distinct from general hidden state dynamicsclaim0.816
  Supported by comparing persona vector transitions to hidden vector transitions from OpenAssistant data
• Persona vectors act as gateway features that gate access to broad repertoires of inferential paths already encoded in the weightsclaim0.811
  Interpretive claim about the mechanistic role of persona vectors explaining emergent misalignment
• There may exist a natural 'persona basis' characterizing the space of all model personas, with correlations between persona vectors predicting co-expression of traitshypothesis0.811
  Open question proposed by authors for future work on the dimensionality and structure of persona space
• The slider metaphor for persona vectors is not the best operationalization; the right one is a mapclaim0.806
  The paper argues that framing persona vectors as uniform dials misses the layered structure of natural/steerable/intractable regimes