Empowering Cross-Domain Sequential Recommendation with Hybrid Tokenization and Serial-Parallel Decoding

Short summary

GenCDSR is a generative framework for cross-domain sequential recommendation that uses hybrid tokenization with a multi-tower architecture to capture both shared and domain-specific patterns. It introduces a serial-parallel decoding strategy that reduces inference latency by 85.1% while improving accuracy by 1.5% over state-of-the-art baselines on three public datasets. Code and datasets are publicly available on GitHub.

• •Hybrid tokenization captures cross-domain commonalities and domain-specific distinctions
• •Serial-parallel decoding cuts inference latency by 85.1% while improving accuracy 1.5%
• •Code and datasets released on GitHub

Generated with AI, which can make mistakes.