CARD: Cluster-level Adaptation with Reward-guided Decoding for Personalized Text Generation

TL;DR

CARD introduces a hierarchical personalization framework for large language models that clusters users and learns individual preferences efficiently during decoding, improving scalability and quality.

Contribution

The paper proposes a novel hierarchical approach combining user clustering, implicit preference learning, and lightweight decoding personalization for scalable user-specific text generation.

Findings

CARD achieves competitive or superior quality on LaMP and LongLaMP benchmarks.

It significantly improves efficiency and scalability for personalized text generation.

The method effectively captures user preferences without manual annotation.

Abstract

Adapting large language models to individual users remains challenging due to the tension between fine-grained personalization and scalable deployment. We present CARD, a hierarchical framework that achieves effective personalization through progressive refinement. CARD first clusters users according to shared stylistic patterns and learns cluster-specific LoRA adapters, enabling robust generalization and strong low-resource performance. To capture individual differences within each cluster, we propose an implicit preference learning mechanism that contrasts user-authored text with cluster-level generations, allowing the model to infer user-specific style preferences without manual annotation. At inference time, CARD injects personalization exclusively at decoding via lightweight user preference vectors and low-rank logit corrections, while keeping the base model frozen. Experiments on the LaMP and LongLaMP benchmarks show that CARD achieves competitive or superior generation quality compared to state-of-the-art baselines, while significantly improving efficiency and scalability for practical personalized text generation.