EUVIMEDOpdag EUVIMED
Open access · CC BY-NC-ND 4.0

Lightweight Denoising and Aligning for Multi-modal Recommender System

Guipeng Xv, Yi Liu, Xinyu Li, Zhenhua Huang, Chen Lin

ACM Transactions on Information Systems · 2026-11-30 · 44 (8) · 1-34

DOI: 10.1145/3847657 ↗

At a glance

Journal article in ACM Transactions on Information Systems, published online on 2026-11-13. Written by Guipeng Xv and 4 co-authors. 59 references. The full text is freely available (open access, CC BY-NC-ND 4.0). Funded by: Natural Science Foundation of China, Fujian Provincial Natural Science Foundation of China, Xiamen Science and Technology Program: Open Call for Proposals Project.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Multi-modal Recommender System (MRS) has emerged as a key information retrieval technology, widely adopted to enhance various web platforms. However, three interrelated challenges remain insufficiently explored: (1) noisy multi-modal content, (2) noisy user feedback, and (3) misalignment between multi-modal content and user feedback. Previous works have either overlooked these challenges or proposed a complex solution. To tackle these challenges in a lightweight way, we propose Lightweight Denoising and Aligning for Multi-modal Recommender System (LDA-MRS). During graph construction, LDA-MRS only constructs a single item–item graph based on consistent cross-modal similarity and dynamic user behavior, effectively reducing noise in multi-modal content. We provide a Lightweight Static Strategy and an Accurate Dynamic Strategy for fusing the graphs. During supervised learning, LDA-MRS leverages multi-modal content to estimate the probability of pairwise observed feedback and introduces a Lightweight Denoising BPR Loss to effectively denoise user feedback. During alignment, LDA-MRS uses Lightweight Alignment guided by User preference to improve task-specific alignment and Lightweight Alignment guided by graded Item relations to achieve finer-grained alignment. LDA-MRS is a lightweight and model-agnostic framework. Experiments on different datasets, backbones, and noisy situations show that LDA-MRS consistently delivers significant performance improvements, highlighting the robustness and effectiveness of LDA-MRS in different conditions.

Funding

  • Natural Science Foundation of China
  • Fujian Provincial Natural Science Foundation of China
  • Xiamen Science and Technology Program: Open Call for Proposals Project

Cite

APA

Xv, G., Liu, Y., Li, X., Huang, Z., & Lin, C. (2026). Lightweight Denoising and Aligning for Multi-modal Recommender System. ACM Transactions on Information Systems, 44(8), 1-34. https://doi.org/10.1145/3847657

Vancouver

Xv G, Liu Y, Li X, Huang Z, Lin C. Lightweight Denoising and Aligning for Multi-modal Recommender System. ACM Transactions on Information Systems. 2026;44(8):1-34. doi:10.1145/3847657

BibTeX
@article{Xv2026,
  author = {Xv, Guipeng and Liu, Yi and Li, Xinyu and Huang, Zhenhua and Lin, Chen},
  title = {Lightweight Denoising and Aligning for Multi-modal Recommender System},
  journal = {ACM Transactions on Information Systems},
  year = {2026},
  volume = {44},
  number = {8},
  pages = {1--34},
  doi = {10.1145/3847657}
}
Start for free →