Mean-difference residual stream injections outperform personality prompting for OCEAN trait steering in most LLMs, with hybrids performing best and showing approximate linearity but non-human trait covariances.
Estimating the Personality of White-Box Language Models
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 2roles
background 1polarities
support 1representative citing papers
LLMs default to responses more similar to opinions from the USA and some European and South American countries; prompting for a country shifts alignment but can introduce stereotypes, while translation does not reliably match language speakers.
citing papers explorer
-
Psychological Steering of Large Language Models
Mean-difference residual stream injections outperform personality prompting for OCEAN trait steering in most LLMs, with hybrids performing best and showing approximate linearity but non-human trait covariances.
-
Towards Measuring the Representation of Subjective Global Opinions in Language Models
LLMs default to responses more similar to opinions from the USA and some European and South American countries; prompting for a country shifts alignment but can introduce stereotypes, while translation does not reliably match language speakers.