From IDs to Semantics: A Generative Framework for Cross-Domain Recommendation with Adaptive Semantic Tokenization

Cross-domain recommendation (CDR) is crucial for improving recommendation accuracy and generalization, yet traditional methods are often hindered by the reliance on shared user/item IDs, which are unavailable in most real-world scenarios. Consequently, many efforts have focused on learning disentangled representations through multi-domain joint training to bridge the domain gaps. Recent Large Language Model (LLM)-based approaches show promise, they still face critical challenges, including: (1) the $\mathbf{\text{item}}\mathbf{\text{ID}}\mathbf{\text{tokenization}}\mathbf{\text{dilemma}}$, which leads to vocabulary explosion and fails to capture high-order collaborative knowledge; and (2) $\mathbf{\text{insufficient}}\mathbf{\text{domain-specific}}\mathbf{\text{modeling}}$ for the complex evolution of user interests and item semantics. To address these limitations, we propose $\mathbf{\text{GenCDR}}$, a novel $\mathbf{\text{Gen}}$erative $\mathbf{\text{C}}$ross-$\mathbf{\text{D}}$omain $\mathbf{\text{R}}$ecommendation framework. GenCDR first employs a $\mathbf{\text{Domain-adaptive}}\mathbf{\text{Tokenization}}$ module, which generates disentangled semantic IDs for items by dynamically routing between a universal encoder and domain-specific adapters. Symmetrically, a $\mathbf{\text{Cross-domain}}\mathbf{\text{Autoregressive}}\mathbf{\text{Recommendation}}$ module models user preferences by fusing universal and domain-specific interests. Finally, a $\mathbf{\text{Domain-aware}}\mathbf{\text{Prefix-tree}}$ enables efficient and accurate generation. Extensive experiments on multiple real-world datasets demonstrate that GenCDR significantly outperforms state-of-the-art baselines. Our code is available in the supplementary materials.