Researchers have introduced OneModel, a unified framework designed to consolidate multiple ranking systems within large-scale platforms like Xiaohongshu. This approach aims to improve user representation and reduce engineering costs by integrating diverse streams such as organic recommendations, advertising, and merchant services. OneModel employs an action-oriented backbone for long-context user representations and a novel Scenario-aware Information Modulation mechanism to balance cross-stream learning with stream-specific needs. Production deployment at Xiaohongshu demonstrated significant online improvements, including a 0.33% increase in Time Spent and a 1.25% rise in Engagement for the Explore Feed, alongside boosts in advertising and merchant recommendation metrics. AI
IMPACT This unified approach to ranking systems could streamline operations and improve user experience across diverse platform services.
RANK_REASON The cluster describes a research paper detailing a new framework for ranking 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 →
