Large language models improve personality control with natural conversation
FATE-5: Faithful Alignment of Trait Expression for Continuous Big Five Personality Control
Computation and Language
Summary
It can be useful for chatbots to have different personalities depending on who they talk to. But it’s hard for computer programs to show strong personality traits while still sounding natural. The authors created a new method called FATE-5 that helps large language models adjust their personality in a smooth, continuous way without losing natural conversation quality. They tested their approach on several models and found it expressed personality traits well while making dialogue sound more natural than other methods.
What this means in practice
- •For chatbot developers: Customize chatbots to express continuous personality traits while maintaining natural conversation for better user interaction.$Commercial implications: Enables building personalized conversational agents for customer service or entertainment that adjust personality continuously.
- •For customer support teams: Use personality-controlled dialogue systems to adapt communication style to different customer moods and preferences.
Authors
Gunhee Cho, Hongyeob Kim, Yun-Gyung Cheong
Abstract
Controlling personality in large language models is important for building conversational agents that can be customized to different users and interaction settings. However, existing personality control methods often face a tradeoff between strong trait expression and natural dialogue generation. We introduce FATE-5, Faithful Alignment of Trait Expression for the Big Five, a two-stage framework that separates personality encoding from naturalness alignment. FATE-5 first learns continuous soft prompt representations of Big Five profiles using dialogue supervision and an auxiliary survey objective. It then improves dialogue naturalness through preference alignment based on both dialogue quality and personality agreement. Across three language model backbones, FATE-5 maintains strong trait expression while achieving substantially higher dialogue naturalness than other trait-strong methods. On held-out questionnaires, it achieves monotonicity correlations of 0.97-0.99 and recovery RMSE of 8.9-12.4, substantially improving over the evaluated baselines. These results demonstrate effective continuous personality control while maintaining natural conversational behavior.