Generative Recommendation for Large-Scale Advertising

2026-02-26

AI Summary

This work proposes GR4AD, a production-ready generative recommendation system for large-scale advertising, addressing the challenges of integrating model architecture, training, and online serving in real-time generative recommendation. Traditional large language models are ill-suited for direct deployment in such settings due to efficiency and scalability constraints. GR4AD innovatively combines unified semantic IDs (UA-SID), lazy autoregressive decoding (LazyAR), value-aware supervised learning (VSL), and ranking-guided soft preference optimization (RSPO), complemented by dynamic beam search for adaptive computational resource allocation. Deployed at full scale on Kuaishou, the system serves over 400 million users, with A/B experiments demonstrating up to a 4.2% increase in ad revenue, while achieving high efficiency, strong scalability, and substantial business impact.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Ad search and search for Web retail
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Generative recommendation has recently attracted widespread attention in industry due to its potential for scaling and stronger model capacity. However, deploying real-time generative recommendation in large-scale advertising requires designs beyond large-language-model (LLM)-style training and serving recipes. We present a production-oriented generative recommender co-designed across architecture, learning, and serving, named GR4AD (Generative Recommendation for ADdvertising). As for tokenization, GR4AD proposes UA-SID (Unified Advertisement Semantic ID) to capture complicated business information. Furthermore, GR4AD introduces LazyAR, a lazy autoregressive decoder that relaxes layer-wise dependencies for short, multi-candidate generation, preserving effectiveness while reducing inference cost, which facilitates scaling under fixed serving budgets. To align optimization with business value, GR4AD employs VSL (Value-Aware Supervised Learning) and proposes RSPO (Ranking-Guided Softmax Preference Optimization), a ranking-aware, list-wise reinforcement learning algorithm that optimizes value-based rewards under list-level metrics for continual online updates. For online inference, we further propose dynamic beam serving, which adapts beam width across generation levels and online load to control compute. Large-scale online A/B tests show up to 4.2% ad revenue improvement over an existing DLRM-based stack, with consistent gains from both model scaling and inference-time scaling. GR4AD has been fully deployed in Kuaishou advertising system with over 400 million users and achieves high-throughput real-time serving.

Problem

• Generative Recommendation
• Large-Scale Advertising
• Real-Time Serving
• Business Value Alignment
• Multi-Candidate Generation

Innovation

• Generative Recommendation
• Lazy Autoregressive Decoding
• Value-Aware Learning
• Ranking-Guided Preference Optimization
• Dynamic Beam Serving