An LLM-native five-factor psychometric instrument produces stable self-report structure but fails to predict observed behavior, and reveals a shared textual-surface bias between self-report and LLM judges that human raters do not share.
Title resolution pending
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3representative citing papers
Big Five inventories fail to capture meaningful differences or recover the five-factor structure in LLMs, with only 3% variance between models and four facets collapsing (r >= .92).
Big Five personality traits become decodable early in LLMs, are represented by mid-layer selective neurons, and can be shifted by targeted activation interventions, though effects on generated labels are weaker and spill across traits.
citing papers explorer
-
An LLM-Native Psychometric Instrument Does Not Predict LLM Behavior: Evidence Across 25 Models
An LLM-native five-factor psychometric instrument produces stable self-report structure but fails to predict observed behavior, and reveals a shared textual-surface bias between self-report and LLM judges that human raters do not share.
-
Personality Without Persons? A Psychometric Critique of Big Five Testing in Large Language Models
Big Five inventories fail to capture meaningful differences or recover the five-factor structure in LLMs, with only 3% variance between models and four facets collapsing (r >= .92).