Tag
This paper introduces a framework for extracting behaviorally grounded user profiles from social media posts to enhance personalized alignment and multi-perspective reasoning in large language models, outperforming synthetic profile methods.