arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2601.06352cs.AI
收藏

CARD: 基于奖励引导解码的集群级适应用于个性化文本生成

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

University of California, Irvine(加州大学尔湾分校);University of Amsterdam(阿姆斯特丹大学);TikTok Inc.(TikTok公司)

Yutong Song, Jiang Wu, Weijia Zhang, Chengze Shen, Shaofan Yuan, Weitao Lu, Jian Wang, Yu Wang, Nikil Dutt, Amir M. Rahmani

AI总结:

CARD通过分层框架实现有效个性化,通过集群用户风格模式聚类和LoRA适配器学习,结合隐式偏好学习机制和解码时的轻量级偏好向量注入,提升个性化文本生成的效率和可扩展性。

摘要:

适应大型语言模型以满足个体用户需求仍具有挑战性,因为细粒度个性化与可扩展部署之间存在张力。我们提出了CARD,一个分层框架,通过逐步细化实现有效个性化。CARD首先根据共享的风格模式对用户进行聚类,并学习集群特定的LoRA适配器,从而实现稳健的泛化和强大的低资源性能。为了在每个集群内捕捉个体差异,我们提出了一种隐式偏好学习机制,通过对比用户自撰文本与集群级生成文本,使模型能够推断用户特定的风格偏好,而无需手动标注。在推理时,CARD通过轻量级用户偏好向量和低秩logit修正,在解码时注入个性化,同时保持基础模型冻结。在LaMP和LongLaMP基准测试中,CARD在生成质量上与最先进的基线相比具有竞争力或更优,同时显著提高了实际个性化文本生成的效率和可扩展性。

英文摘要:

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 group-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 superior generation quality compared to baselines, while significantly improving efficiency and scalability for practical personalized text generation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑