Cross-Scenario Unified Modeling of User Interests at Billion Scale

Manjie Xu, Cheng Chen, Xin Jia, Jingyi Zhou, Yongji Wu, Wang Zejian, Chi Zhang, Kai Zuo, Nemo Chen, Xu Tang, Yao Hu, Yixin Zhu

Keywords: sequential recommendation, behavior modeling, multi-scenario reasoning

TL;DR: This paper presents a unified hierarchical recommender system leveraging large language models and multi-scenario behavior modeling for more accurate and scalable content recommendations.

Abstract: User interests on content platforms are inherently diverse, manifesting through complex behavioral patterns across heterogeneous scenarios such as search, feed browsing, and content discovery. Traditional recommendation systems typically prioritize business metric optimization within isolated specific scenarios, neglecting cross-scenario behavioral signals and struggling to integrate advanced techniques like LLMs at billion-scale deployments, which finally limits their ability to capture holistic user interests across platform touchpoints. We propose **RED-Rec**, an LLM-enhanced hierarchical **R**ecommender **E**ngine for **D**iversified scenarios, tailored for industry-level content recommendation systems. RED-Rec unifies user interest representations across multiple behavioral contexts by aggregating and synthesizing actions from varied scenarios, resulting in comprehensive item and user modeling. At its core, a two-tower LLM-powered framework enables nuanced, multifaceted representations with deployment efficiency, and a scenario-aware dense mixing and querying policy effectively fuses diverse behavioral signals to capture cross-scenario user intent patterns and express fine-grained, context-specific intents during serving. We validate RED-Rec on hundreds of millions of users in a world-leading UGC platform through online A/B testing, showing substantial performance gains in both content recommendation and advertisement targeting tasks. We further introduce a million-scale sequential recommendation dataset for comprehensive offline training and evaluation. We hope our work could advance unified modeling of users, unlocking deeper personalization and fostering more meaningful user engagement across large-scale platforms.

Primary Area: learning on time series and dynamical systems

Submission Number: 1947