Hierarchical Quantization with Domain-Adaptive Sparse Routing for Generative Cross-Domain Recommendation

2026-08-07

AI Summary

Existing unified generative models struggle to effectively capture the heterogeneous item semantics and user behaviors inherent in cross-domain recommendation. To address this limitation, this work proposes HD-Rec, a novel framework that employs a hierarchical domain-aware quantizer to generate semantic IDs and introduces a domain-adaptive sparse Mixture-of-Experts (MoE) module to flexibly balance shared and domain-specific representations. Furthermore, the framework incorporates a cross-granularity routing consistency regularization to enhance multi-granularity semantic alignment. Extensive experiments demonstrate that HD-Rec significantly outperforms state-of-the-art sequential, generative, and cross-domain recommendation methods on three public cross-domain benchmark datasets.

Abstract

Generative Recommendation (GenRec) represents a promising paradigm that achieves remarkable empirical success by encoding items as compact Semantic IDs (SIDs) and modeling user behavior via next-token prediction across diverse recommendation scenarios. Extending this paradigm to cross-domain recommendation is challenging because a unified model must accommodate heterogeneous item semantics and behavioral patterns across domains. Existing methods commonly rely on globally shared representations or lightweight domain adaptation, which may provide insufficient capacity for modeling heterogeneous patterns at different semantic granularities. To address these challenges, we propose HD-Rec, a unified generative framework for cross-domain recommendation. HD-Rec employs a hierarchical domain-aware quantizer that constructs semantic identifiers using globally shared coarse-level codebooks and adaptively routed fine-level codebooks. It further introduces a domain-adaptive sparse mixture-of-experts module that combines a continuously activated shared expert with a dynamically selected specialized expert. To improve the coherence of multi-token item representations, we develop a cross-granularity routing consistency objective that regularizes token-level routing decisions toward their item-level consensus. Experiments on three public cross-domain recommendation benchmarks show that HD-Rec consistently improves over competitive sequential, generative, and cross-domain recommendation baselines.

Problem

• cross-domain recommendation
• generative recommendation
• heterogeneous semantics
• semantic granularity
• domain adaptation

Innovation

• Hierarchical Quantization
• Domain-Adaptive Sparse Routing
• Generative Recommendation
• Cross-Domain Recommendation
• Mixture-of-Experts