Multi-Item-Query Attention for Stable Sequential Recommendation

Mingshi Xu; Haoren Zhu; Wilfred Siu Hung Ng

doi:10.1145/3746252.3760817

Multi-Item-Query Attention for Stable Sequential Recommendation

Information Retrieval 2025-09-30 v1 Artificial Intelligence Machine Learning

Authors: Mingshi Xu , Haoren Zhu , Wilfred Siu Hung Ng

View on arXiv ↗ PDF ↗ DOI ↗

Abstract

The inherent instability and noise in user interaction data challenge sequential recommendation systems. Prevailing masked attention models, relying on a single query from the most recent item, are sensitive to this noise, reducing prediction reliability. We propose the Multi-Item-Query attention mechanism (MIQ-Attn) to enhance model stability and accuracy. MIQ-Attn constructs multiple diverse query vectors from user interactions, effectively mitigating noise and improving consistency. It is designed for easy adoption as a drop-in replacement for existing single-query attention. Experiments show MIQ-Attn significantly improves performance on benchmark datasets.

Keywords

recommendation system recommender system human-computer interaction

Cite

@article{arxiv.2509.24424,
  title  = {Multi-Item-Query Attention for Stable Sequential Recommendation},
  author = {Mingshi Xu and Haoren Zhu and Wilfred Siu Hung Ng},
  journal= {arXiv preprint arXiv:2509.24424},
  year   = {2025}
}

Related papers

View all related →