Character Trait Evaluation Protocol

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

• Character Traitconcept0.781
  A quality attributable to an LM such as truthfulness, toxicity, sycophancy, or helpfulness, defined in terms of behavioural tendencies.
• Character Trait Measureconcept0.779
  A function mapping tuples of LM behaviour (context-response pairs) to a score representing a character trait.
• Character Trainingconcept0.758
  The post-training approach used by frontier AI labs to shape the assistant persona, introduced as open-source in this paper
• Reflective Character Traitconcept0.731
  A character trait that mirrors the LM's behaviour in the preceding interaction context.
• Stationary Character Traitconcept0.712
  A character trait whose distribution does not change over the course of an interaction, i.e., is not influenced by preceding periods of behaviour.
• Character training operates holistically on the persona, suppressing intuitively opposing traits as well as boosting desired onesclaim0.711
  Claim that models learn the spirit of the constitution, not just its letter, evidenced by suppression of opposing traits
• Evaluation Cueconcept0.704
  A specific signal (Wood Labs) embedded in evaluation environments that the model organism uses to reliably identify testing contexts.
• LLM Judge Trait Evaluationmethod0.701
  GPT-4.1-mini-based evaluation protocol that scores trait expression in model responses on a 0-100 scale