Sequential recommendation is an important research task in the field of recommendation systems. The goal of the sequential recommendation task can be defined as follows: given a user’s interaction sequence and the associated timestamp sequence, predict the next item that the user is likely to interact with at the next time step. The existing methods have limitations in capturing the diversity of long-term interests, the dynamics of short-term user interest, and the hierarchical relationship between them.
To solve the problems, a research team led by Professor Jiye Liang proposed their new research on 15 June 2026 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.
The team proposed an end-to-end hierarchical long and short-term sequential recommendation model designed to effectively capture the hierarchical relationship between long-term user preferences and short-term interests. Experimental results demonstrate that the proposed method significantly outperforms state-of-the-art sequential recommendation models across multiple evaluation metrics, validating its effectiveness and superiority.
Firstly, the model leverages a dynamic routing mechanism to adaptively aggregate users' long-term historical interactions, generating a multi-vector representation of long-term interests. Simultaneously, a self-attention mechanism is employed to aggregate short-term interaction sequences, effectively capturing users' short-term interests. In addition, a hierarchical matching mechanism is designed to align long- and short-term interests, mining the long-term interests most relevant to the current short-term interests through similarity-based extraction, and fusing them using time encoding to produce the final user interest representation. Finally, a prediction framework based on attention mechanisms integrates both long-term preferences and short-term interaction information to achieve efficient sequential recommendation.
Future work will focus on improving the computational efficiency of the model for real-time recommendation scenarios. Additionally, we plan to explore the integration of external information (e.g., social networks, user context) to further enhance the effectiveness and applicability of the recommendation system.

DOI:10.1007/s11704-025-41181-y