Researchers have developed a new framework for sponsored product ranking in search engines that improves upon existing user sequence modeling techniques. This framework incorporates dual retrieval channels over long-term user history, specifically candidate-nearest-neighbor and query-nearest-neighbor retrieval. By encoding these retrieved actions alongside query and candidate context, and using candidate-aware early fusion, the model significantly boosts key ranking metrics. Online A/B testing demonstrated increases in click-through rate, attributed conversion rate, and orders per impression. AI
IMPACT This research could lead to more effective sponsored product ranking, improving user experience and conversion rates in search engines.
RANK_REASON The item describes a research paper detailing a new model for sponsored product ranking in search. [lever_c_demoted from research: ic=1 ai=0.7]
Read on Hugging Face Daily Papers →
- click-through rate
- Hugging Face
- Query- and Candidate-Aware Long User Sequence Modeling for Sponsored Product Ranking in Search
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →