Some personality traits may be less accessible using linear methods than others

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

• Is there a tradeoff between subtlety of trait expression and robustness in character-trained models?question0.766
  Open question raised in Appendix E regarding misaligned persona behavior
• What is the exact mechanism by which synthetic introspective data improves character trait expression — how does varying amount, diversity, or source affect outcomes?question0.760
  Open question identified in Discussion as future work
• What factors determine the consistency with which LMs exhibit character traits?question0.759
  Empirical question answered through experiments on model size, fine-tuning, and prompting.
• Almost any social process can have a relatively more living character, or have a relatively less living character.claim0.756
• What distinguishes personas from traits, and do emergence and persistence patterns carry over to compositional or character-level constructs?question0.755
  Conceptual open question about formalizing the persona abstraction beyond the trait level
• The consistency with which LMs exhibit character traits varies with model size, fine-tuning, and prompting.claim0.754
  Finding replicated across multiple experiments.
• What was difficult or impossible as a larger act of social transformation becomes possible when one uses a genetic approach of small snippets.claim0.753
  Summarizes the practical advantage of breaking complex new processes into small, individually adoptable sequences.
• Linear truth directions in LLMs are reliable primarily in factual recall cases and break down when truth assessment depends on computing and storing intermediate results.claim0.751
  Central empirical conclusion of the paper about the fundamental limits of truth directions.