Editing Personality for LLMs
This task seeks to adjust the models’ responses to opinion-related questions on specified topics since an individual’s personality often manifests in the form of their expressed opinions, thereby showcasing different personality traits.
Abstract
This paper introduces an innovative task focused on editing the personality traits of Large Language Models (LLMs). This task seeks to adjust the models’ responses to opinion-related questions on specified topics since an individual’s personality often manifests in the form of their expressed opinions, thereby showcasing different personality traits. Specifically, we construct a new benchmark dataset PersonalityEdit to address this task. Drawing on the theory in Social Psychology (Goldberg, 1990), we isolate three representative traits, namely N EUROTICISM , E XTRAVERSION