Measuring What Makes You Unique: Difference-Aware User Modeling for Enhancing LLM Personalization

Yilun Qiu^1*, Xiaoyan Zhao^2*, Yang Zhang^1†, Yimeng Bai³, Wenjie Wang³, Hong Cheng², Fuli Feng³, Tat-Seng Chua¹

¹National University of Singapore, ²The Chinese University of Hong Kong, ³University of Science and Technology of China

^*Equal Contribution, ^†Corresponding author

This is the implementation of the Difference-aware Personalization Learning (DPL) method proposed in our paper accepted by ACL 2025 Findings.

📋 Catalogue

• 📋 Catalogue
• ⚙️ Environment Setup
• 📚 Dataset Preprocess
• ⌛️ Quick Start
• 📊 Experimental Results
• 📖 Citation

⚙️ Environment Setup

conda create -n DPL python=3.11.11
conda activate DPL
pip install -r requirements.txt

📚 Dataset Preprocess

The dataset we used in DPL is adapted from Amazon Reviews'23. We publish our processed datasets in Huggingface. You can also process the dataset yourself and store it locally by the following commands:

cd data/
./create.sh

⌛️ Quick Start

To execute the DPL method, please first complete the required information in the .env file. Then, run the following command:

./main.sh

You can modify the main.sh file to change parameters.

📊 Experimental Results

📖 Citation

If you find our work useful, please kindly cite our paper:

@article{qiu2025measuring,
  title={Measuring What Makes You Unique: Difference-Aware User Modeling for Enhancing LLM Personalization},
  author={Qiu, Yilun and Zhao, Xiaoyan and Zhang, Yang and Bai, Yimeng and Wang, Wenjie and Cheng, Hong and Feng, Fuli and Chua, Tat-Seng},
  journal={arXiv preprint arXiv:2503.02450},
  year={2025}
}