A Survey on Robust Sequential Recommendation: Fundamentals, Challenges, Taxonomy, and Future Directions
ACM Transactions on Information Systems · 07.11.2026
DOI: 10.1145/3860537 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Information Systems, online erschienen am 07.11.2026. Verfasst von Yatong Sun und 3 weiteren Autor:innen. Die Arbeit nennt 209 Literaturangaben.
Abstract (Originalsprache)
Sequential recommender systems (SRSs) have become essential for modeling users’ evolving preferences and supporting personalized information filtering. Most existing SRSs rely on the assumption that the input and target of each data instance are sequentially relevant. However, this assumption is frequently violated because real-world interactions arise from complex hidden motivations, including personal preferences, behavioral randomness, contextual influences, and malicious attacks. Motivational inconsistency among interactions can create spurious sequential patterns, forming unreliable instances with mismatched input–target pairs that mislead model training and inference and ultimately impair recommendation accuracy. To mitigate unreliable instances, R obust S equential R ecommenders (RSRs) have emerged as an important research direction. This survey presents the first systematic review of RSR research. We first analyze unreliable instances in sequential recommendation, clarifying their causes, manifestations, and adverse impacts. We then identify the distinctive challenges of RSRs, highlighting how they differ from robustness problems in non-sequential recommendation and from label denoising tasks in computer vision and natural language processing. Building on these foundations, we propose a lifecycle-oriented taxonomy that organizes existing RSRs into architecture-centric, data-centric, learning-centric, and inference-centric paradigms. We further conduct a comparative analysis of representative RSRs along seven key properties, revealing their strengths and weaknesses. In addition, we summarize commonly used evaluation metrics and benchmarks for RSRs. Finally, we discuss open issues and promising future directions. To facilitate further research, we maintain a curated repository of RSR resources at https://github.com/AlchemistYT/Awesome-RSRs .
Zitieren
APA
Sun, Y., Yang, X., Wang, B., & Sun, Z. (2026). A Survey on Robust Sequential Recommendation: Fundamentals, Challenges, Taxonomy, and Future Directions. ACM Transactions on Information Systems. https://doi.org/10.1145/3860537
Vancouver
Sun Y, Yang X, Wang B, Sun Z. A Survey on Robust Sequential Recommendation: Fundamentals, Challenges, Taxonomy, and Future Directions. ACM Transactions on Information Systems. 2026. doi:10.1145/3860537
BibTeX
@article{Sun2026,
author = {Sun, Yatong and Yang, Xiaochun and Wang, Bin and Sun, Zhu},
title = {A Survey on Robust Sequential Recommendation: Fundamentals, Challenges, Taxonomy, and Future Directions},
journal = {ACM Transactions on Information Systems},
year = {2026},
doi = {10.1145/3860537}
}
EUVIMED