Researchers have developed POEM (Partial-Order Enhanced Modeling), a novel framework for real-time sequential modeling in recommendation systems. This approach addresses the limitations of traditional models by incorporating dynamic user interests and contextual conditions. POEM leverages partial-order relations derived from multi-stage ranking scores, such as predicted click-through rates and watch durations, to construct richer sequences. When deployed on Kuaishou's online traffic, POEM demonstrated significant improvements, increasing average per-user watch time by approximately 0.249% on the KS Single Page and 0.213% on the KS Lite Page. AI
IMPACT Enhances real-time recommendation systems by better capturing dynamic user interests and optimizing for multiple ranking targets.
RANK_REASON Academic paper detailing a new modeling framework for recommendation systems. [lever_c_demoted from research: ic=1 ai=0.7]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
