DOS: Dual-Flow Orthogonal Semantic IDs for Recommendation in Meituan

2026-02-04

AI Summary

This work addresses the limitations of existing semantic ID methods in recommender systems, which lack contextual awareness and suffer from misalignment between the semantic codebook space and the generation space, leading to significant semantic loss due to suboptimal quantization in large language models. To mitigate these issues, we propose Dual-stream Orthogonal Semantic IDs (DOS), a novel approach that integrates collaborative signals via a user-item dual-stream framework to align semantic spaces and introduces an orthogonal residual quantization mechanism that rotates the semantic space to maximally preserve semantic information. By synergistically combining dual-stream collaborative modeling with orthogonal residual vector quantization, DOS effectively bridges the semantic gap and reduces information loss. Extensive offline experiments and online A/B tests demonstrate its substantial performance gains, and the method has been deployed in the Meituan app, serving hundreds of millions of users.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Search and Optimization: Distributed Search
• Machine Learning: Semi-Supervised Learning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Semantic IDs serve as a key component in generative recommendation systems. They not only incorporate open-world knowledge from large language models (LLMs) but also compress the semantic space to reduce generation difficulty. However, existing methods suffer from two major limitations: (1) the lack of contextual awareness in generation tasks leads to a gap between the Semantic ID codebook space and the generation space, resulting in suboptimal recommendations; and (2) suboptimal quantization methods exacerbate semantic loss in LLMs. To address these issues, we propose Dual-Flow Orthogonal Semantic IDs (DOS) method. Specifically, DOS employs a user-item dual flow-framework that leverages collaborative signals to align the Semantic ID codebook space with the generation space. Furthermore, we introduce an orthogonal residual quantization scheme that rotates the semantic space to an appropriate orientation, thereby maximizing semantic preservation. Extensive offline experiments and online A/B testing demonstrate the effectiveness of DOS. The proposed method has been successfully deployed in Meituan's mobile application, serving hundreds of millions of users.

Problem

• Semantic IDs
• contextual awareness
• quantization
• semantic loss
• recommendation systems

Innovation

• Semantic IDs
• Dual-Flow Framework
• Orthogonal Quantization
• Generative Recommendation
• Collaborative Signals