DSRec Cross-SSM Model for Dual-Interest Sequential Recommendation
September 21, 2026
DSRec uses a dual-interest cross-State Space Model (SSM) architecture to disentangle long-term and short-term item semantic roles. The model addresses item polysemy and improves cost efficiency over Transformer-based sequential recommendation methods.
HOW THIS AFFECTS YOU
●
builderYou can implement more granular user interest modeling while reducing the computational cost of Transformers.
●
researcherYou can explore SSM-based architectures for more efficient long-range sequential modeling.