Reproduce “A Personalized Dense Retrieval Framework for Unified Information Access” 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
Hansi Zeng, Surya Kallumadi, Zaid Alibadi, Rodrigo Nogueira, Hamed Zamani
A unified dense retrieval framework, UIA, is introduced to handle diverse information access functionalities like keyword search, query by example, and complementary item recommendation within a single model. The framework incorporates an Attentive Personalization Network (APN) to adapt results based on user history, consistently outperforming specialized baselines across tasks on large-scale e-commerce datasets.
Key insights
- A single dense retrieval model can effectively unify diverse information access tasks, fostering knowledge transfer and improved generalization across functionalities.
- Explicitly encoding the information access functionality within the request is crucial for task differentiation and performance in a unified framework.
- Personalization, particularly through an Attentive Personalization Network (APN) that leverages cross-task user history, provides substantial performance gains, especially for tasks with less explicit user intent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.