MoF method personalizes black box AI models efficiently
MoF: Preference-Aware Mixture Modeling for Black-Box LLM Personalization
Artificial Intelligence
Summary
It is hard to make large AI language models match what different people want because each user can have unique preferences. The authors introduce a way to personalize these AI models without needing separate custom settings for each user, which usually gets very complicated. Their method, called Mixture-of-Facets (MoF), combines shared pieces that represent user preferences, allowing it to work well even for new users it has never seen before. This approach makes personalization more scalable and uses fewer extra parameters than older methods.
What this means in practice
- •For software developers: Customize black-box large language models to individual users efficiently without extra per-user training.
- •For customer support platforms: Enable AI chatbots to better align with diverse client preferences without costly retraining for each customer.$Commercial implications: Allows offering tailored AI chat services to many clients with less computational cost, improving service personalization.
Authors
Hun Park
Abstract
Proprietary Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet aligning their outputs with diverse user preferences remains challenging. Existing personalization approaches for black-box LLMs often rely on user-specific scoring heads, causing the number of personalized parameters to grow linearly with the number of users and requiring additional adaptation for unseen users. To address these limitations, we propose Mixture-of-Facets (MoF), a scalable personalization framework for black-box LLMs that models user preferences as compositions of shared latent preference facets rather than dedicated user-specific parameters. MoF performs personalization through history-conditioned routing over shared facet heads, enabling personalization for users unseen during training without additional parameter updates. Across diverse personalization tasks, MoF delivers stronger personalization performance while maintaining a more scalable and parameter-efficient design than prior approaches. Additional analysis indicates strong generalization to unseen users.