Large language models personalize quickly with tiny user data
Unlocking Latent Personalization in LLMs
Artificial Intelligence
Summary
Personalizing large language models (LLMs) to match individual users is usually hard when there is very little user data. The authors suggest that LLMs already have hidden abilities to adapt, so just a few examples from a user can help guide the model to behave more personally without heavy retraining. They created a method called LatentPersonal that represents a user’s style in a small, easy-to-learn format and uses that to adjust the model quickly. Tests show it helps models personalize well even from just one example.
What this means in practice
- •For chatbot developers: Enable chatbots to quickly custom-tailor interactions to individual users with very limited data input.
- •For customer support platform teams: Reduce the workload for customer service AI by swiftly adjusting language models to the tones and preferences of different clients.
Authors
Wei Chen, Guanghui Zhu, Zhongliang Cai, Yihua Huang
Abstract
Large language models (LLMs) are increasingly expected to adapt to individual users, yet effective personalization remains challenging when only limited user-specific samples are available. In this work, we take an alternative perspective: pretrained LLMs may already possess latent capacity for personalization, and a few user samples may therefore suffice to guide the model toward user-aligned behavior with minimal user-specific adaptation. From this perspective, we propose LatentPersonal, a framework that formulates personalization as navigation in a shared latent adaptation space. LatentPersonal infers a compact latent representation from a few user samples to guide user-specific model adaptation, regularized with a variational information bottleneck to encourage compact preference representations. We instantiate LatentPersonal with LoRA, leveraging its low-rank parameterization as a natural low-dimensional adaptation space for personalization. By simply inserting a user-specific guidance vector between the shared low-rank factors, the model can navigate toward personalized adaptations through lightweight inference of this compact representation, without updating the shared LoRA parameters. Experiments across multiple personalization datasets demonstrate that LatentPersonal substantially reduces user-specific adaptation overhead while achieving effective personalization from only a few user-specific interactions, with particularly strong performance in the one-shot regime.