Mosaic: A Fleet of User Embedding Specialists for Recommendation at Meta

2026-07-27

AI Summary

This work proposes a hybrid architecture comprising four heterogeneous expert models—memory-driven, dense-dominant, sequential, and co-trained—to efficiently construct highly expressive user representations for large-scale recommendation systems, effectively supporting retrieval, ranking, and integrity tasks. The approach comprehensively captures multidimensional user behavioral patterns and introduces two novel mechanisms: Marginal Representation Maximization (MRM) and Collaborative Representation Learning (CRL), which enhance the marginal information contribution of each expert. To accelerate iterative development, the authors design CoEval and User Tower Zero-Out—log-free evaluation frameworks that enable rapid model assessment. The system integrates memory networks, dense models, sequential modeling, and collaborative training within a hybrid CPU/GPU online/offline serving infrastructure, allowing flexible embedding updates and inference strategies. Experiments demonstrate a significant reduction in normalized entropy (NE) and consistent, substantial gains in live recommendation performance.

Abstract

User representation is one of the highest-leverage modeling problems in industrial recommendation systems: a single advancement in how users are encoded can propagate across retrieval, ranking, and integrity tasks at platform scale. Prior industrial user representation work builds either a single user model that emits one or more embedding vectors or a shared backbone with task-specific adaptation. In this paper, we present Mosaic, a foundational user modeling platform that employs a fleet of specialists to learn user embeddings. The fleet comprises four architecturally diverse model families - memorization-driven, dense-heavy, sequential-based, and CoTrain models - each focusing on a distinct facet of user behavior. We developed MRM (Multi-task Relations Mining) and CRL (Cosine Redundancy Loss) techniques to maximize the marginal information contribution of each new specialist. We also introduce CoEval and User Tower Zero-Out, new logging-free embedding evaluation framework that improves development velocity while preserving downstream-aligned accuracy. Our hybrid CPU/GPU, online-and-offline serving stack allows each specialist to choose the adequate serving strategy to meet the freshness, latency, and computational requirements. Mosaic delivers consistent and significant offline NE improvements in addition to online gains.

Problem

• user representation
• recommendation systems
• user embedding
• industrial recommendation
• foundational user modeling

Innovation

• user representation
• model specialization
• multi-task learning
• embedding evaluation
• hybrid serving