What factors determine the consistency with which LMs exhibit character traits?

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

• The consistency with which LMs exhibit character traits varies with model size, fine-tuning, and prompting.claim0.919
  Finding replicated across multiple experiments.
• How do character traits develop over the course of an LM interaction?question0.823
  Core empirical question motivating Section 5 on stationary and reflective traits.
• LMs can consistently exhibit more or less accurate and logically coherent beliefs on specific input distributions.claim0.822
  Conclusion from Experiment 2 on Leap-of-Thought.
• LM Character Traits Frameworkframework0.784
  The paper's central contribution: a formal behaviourist framework for attributing character traits to LMs based on input-output behaviour.
• Whether LM question-answering consistency constitutes evidence for beliefs in a meaningful sense depends on the behaviourist operationalisation chosen.claim0.779
  Epistemic caveat the authors use to qualify their empirical findings on LM beliefs.
• We hypothesize that LLMs represent correctness of arithmetic expressions differently from factual statements.hypothesis0.779
  Core working hypothesis motivating the factual vs. arithmetic task split in the experimental design.
• Personas matter for LLM individuation because they exhibit stable beliefs, values and traits and provide a new class of individuation targetsclaim0.772
  Central thesis of the paper's second contribution connecting persona research to the individuation problem
• Automated interpretability using LLMs can usefully score feature specificity.claim0.768
  Claude 3 Opus ratings aligned with human judgment of feature descriptions.