cs.AIJun 1, 2026

From Profiles to Steering Vectors: Global Sparse Priors and Local Semantic Calibration for Personalized Text Generation

Authors: Liuji Chen, Zeyu Zhang, Xinyuan Zhang, Shuai Nie, Qiang Liu, Shu Wu, Liang Wang

Organizations: NLPR, Institute of Automation, Chinese Academy of Sciences · Xiaomi

Abstract

Personalized text generation requires models to capture user-specific writing styles from historical data. Existing approaches based on retrieval, parameter-efficient fine-tuning, or activation steering either introduce inference and storage overhead or struggle to separate stylistic signals from semantic content. We propose GLASS, a training-free framework for personalized generation via Global-Local Activation Steering with Sparse priors. GLASS uses sparse autoencoders to extract a global user-style prior from historical responses and constructs local contrastive style vectors over clustered interaction scenarios. During inference, it jointly injects global and local vectors into different model layers, enabling context-aware personalization without retrieval or parameter updates. Experiments on LaMP and LongLaMP show that GLASS outperforms retrieval-, fine-tuning-, and steering-based baselines across ROUGE metrics and LLM-as-judge evaluations. Further analyses show that SAE-based representations are more robust to topic and length shifts, suggesting better disentanglement of stylistic information from semantic residue.

Figures & tables

Appendix figures & tables17 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. PrefReward: Learning User Preference Matrix for Personalized Text Generation

    Jul 23, 2026Yue Wu, Chengbing Wang, Yimeng Bai +3Personalized Language ModelsPersonalized Text Generation

  2. Privacy Personalization Trade offs in LLMs: The Impact of Stylometric Signal Reduction on User-Specific Text Generation

    Aug 18, 2026Muhammed Nazmul Arefin, Omar Jamal HammadPersonalized Text GenerationPersonalized Language Models

  3. Iterative Critique-Refine Framework for Enhancing LLM Personalization

    Oct 28, 2025Durga Prasad Maram, Dhruvin Gandhi, Zonghai Yao +5LLM PersonalizationPersonalized Language Models