We hypothesize that axes of persona differentiation within LLMs are likely already present in base models and inherited from the pre-training corpus

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

• There may exist a natural 'persona basis' characterizing the space of all model personas, with correlations between persona vectors predicting co-expression of traitshypothesis0.812
  Open question proposed by authors for future work on the dimensionality and structure of persona space
• Persona representations originate in pretraining, not alignment; they form without explicit supervision as useful features for next-token predictionclaim0.807
  Core interpretive claim providing mechanistic explanation for early persona formation
• How does different post-training data shift a model's position along persona dimensions?question0.800
  Future work direction: using persona space to study effects of training data on model character
• Hypothesis 1 (Gateway Features): Persona vectors act as gateway features—single directions in activation space that shape LLM behavior across most, if not all, contextshypothesis0.792
  First of three hypotheses about persona implementation in LLMs, motivating the persona views
• Persona vectors extracted from base pretraining checkpoints steer fully post-trained OLMo-3-7B-Instructfinding0.792
  Shows persona directions persist through all alignment stages, answering key part of RQ2
• Personas matter for LLM individuation because they exhibit stable beliefs, values and traits and provide a new class of individuation targetsclaim0.790
  Central thesis of the paper's second contribution connecting persona research to the individuation problem
• Persona vector-based data filtering and LLM judge-based data filtering have complementary strengths for identifying problematic training dataclaim0.789
  Author's interpretive conclusion from comparing filtering strategies
• Post-training elicits rather than installs personas; the persona direction is already laid down in pretrainingclaim0.788
  Authors' interpretive endorsement of PSM view, backed by transfer experiments