Scaling Personality Control in LLMs with Big Five Scaler Prompts
Big5-Scaler is a prompt-based method to control Large Language Models' personality using Big Five traits without retraining. It embeds numeric trait values into natural language prompts, offering fine-grained control over personality expression during inference. Evaluation shows it effectively induces consistent and distinguishable traits across models, with simpler prompts and lower intensity levels proving most efficient.