Anchored Alignment: Preventing Positional Collapse in Multimodal Recommender Systems

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Jeong, Yonghun, Kang, David Yoon Suk, Lee, Yeon-Chang
Format:	Preprint
Published:	2026
Subjects:	Information Retrieval Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866910051662299136
author	Jeong, Yonghun Kang, David Yoon Suk Lee, Yeon-Chang
author_facet	Jeong, Yonghun Kang, David Yoon Suk Lee, Yeon-Chang
contents	Multimodal recommender systems (MMRS) leverage images, text, and interaction signals to enrich item representations. However, recent alignment based MMRSs that enforce a unified embedding space often blur modality specific structures and exacerbate ID dominance. Therefore, we propose AnchorRec, a multimodal recommendation framework that performs indirect, anchor based alignment in a lightweight projection domain. By decoupling alignment from representation learning, AnchorRec preserves each modality's native structure while maintaining cross modal consistency and avoiding positional collapse. Experiments on four Amazon datasets show that AnchorRec achieves competitive top N recommendation accuracy, while qualitative analyses demonstrate improved multimodal expressiveness and coherence. The codebase of AnchorRec is available at https://github.com/hun9008/AnchorRec.
format	Preprint
id	arxiv_https___arxiv_org_abs_2603_12726
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Anchored Alignment: Preventing Positional Collapse in Multimodal Recommender Systems Jeong, Yonghun Kang, David Yoon Suk Lee, Yeon-Chang Information Retrieval Machine Learning Multimodal recommender systems (MMRS) leverage images, text, and interaction signals to enrich item representations. However, recent alignment based MMRSs that enforce a unified embedding space often blur modality specific structures and exacerbate ID dominance. Therefore, we propose AnchorRec, a multimodal recommendation framework that performs indirect, anchor based alignment in a lightweight projection domain. By decoupling alignment from representation learning, AnchorRec preserves each modality's native structure while maintaining cross modal consistency and avoiding positional collapse. Experiments on four Amazon datasets show that AnchorRec achieves competitive top N recommendation accuracy, while qualitative analyses demonstrate improved multimodal expressiveness and coherence. The codebase of AnchorRec is available at https://github.com/hun9008/AnchorRec.
title	Anchored Alignment: Preventing Positional Collapse in Multimodal Recommender Systems
topic	Information Retrieval Machine Learning
url	https://arxiv.org/abs/2603.12726