Reproduce “Measuring What Makes You Unique: Difference-Aware User Modeling for Enhancing LLM Personalization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yilun Qiu, Xiaoyan Zhao, Yang Zhang, Yimeng Bai, Wenjie Wang, Hong Cheng, Fuli Feng, Tat-Seng Chua
A new method for Large Language Model personalization, Difference-aware Personalization Learning (DPL), leverages inter-user comparative analysis to model a user's unique preferences by distinguishing them from others. This approach achieved state-of-the-art performance in personalized review generation, with an average relative improvement of 4.29% over best baselines on the CDs & Vinyl dataset, particularly benefiting users with more unique characteristics.
Key insights
- Incorporating inter-user differences into personalization models, rather than just isolated user history, yields more distinct and satisfactory LLM outputs.
- A selective user comparison strategy using clustering effectively identifies diverse comparison groups, providing a global perspective on user uniqueness without extensive computation.
- Structuring difference extraction by focusing on specific preference dimensions, such as writing, emotional, and semantic styles, leads to more relevant and actionable personalization cues for LLMs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.