Cross-Scenario Unified Modeling of User Interests at Billion Scale

2025-10-16

AI Summary

Traditional recommender systems optimize for single scenarios, neglecting cross-scenario behavioral synergy and struggling to integrate large language models (LLMs) at billion-scale, leading to fragmented user interest modeling. To address this, we propose RED-Rec—a novel framework that pioneers deep LLM integration into industrial-scale recommendation systems. It introduces a dual-tower LLM architecture coupled with a scenario-aware dense hybrid query mechanism, enabling hierarchical fusion of multi-scenario behavioral sequences (e.g., search, feed, content discovery) and fine-grained interest modeling. An efficient online serving engine supports low-latency real-time inference. Extensive A/B tests across hundreds of millions of users demonstrate significant improvements in core recommendation and advertising metrics. Furthermore, we release RED-MMU—a million-scale, multi-scenario sequential dataset—to foster collaborative advancement in both academia and industry.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

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 Recommender Engine for Diversified 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 through online A/B testing on hundreds of millions of users in RedNote 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, RED-MMU, for comprehensive offline training and evaluation. Our work advances unified user modeling, unlocking deeper personalization and fostering more meaningful user engagement in large-scale UGC platforms.

Problem

• Modeling diverse user interests across multiple scenarios
• Integrating LLMs efficiently in billion-scale recommendation systems
• Unifying behavioral signals from heterogeneous platform interactions

Innovation

• LLM-enhanced hierarchical recommender for diverse scenarios
• Two-tower LLM framework enables efficient multifaceted representations
• Scenario-aware policy fuses behavioral signals across contexts