Decoupled Multimodal Fusion for User Interest Modeling in Click-Through Rate Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fan, Alin, Li, Hanqing, Lu, Sihan, Yuan, Jingsong, Zhang, Jiandong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917412488609792
author Fan, Alin
Li, Hanqing
Lu, Sihan
Yuan, Jingsong
Zhang, Jiandong
author_facet Fan, Alin
Li, Hanqing
Lu, Sihan
Yuan, Jingsong
Zhang, Jiandong
contents Modern industrial recommendation systems improve recommendation performance by integrating multimodal representations from pre-trained models into ID-based Click-Through Rate (CTR) prediction frameworks. However, existing approaches typically adopt modality-centric modeling strategies that process ID-based and multimodal embeddings independently, failing to capture fine-grained interactions between content semantics and behavioral signals. In this paper, we propose Decoupled Multimodal Fusion (DMF), which introduces a modality-enriched modeling strategy to enable fine-grained interactions between ID-based collaborative representations and multimodal representations for user interest modeling. Specifically, we construct target-aware features to bridge the semantic gap across different embedding spaces and leverage them as side information to enhance the effectiveness of user interest modeling. Furthermore, we design an inference-optimized attention mechanism that decouples the computation of target-aware features and ID-based embeddings before the attention layer, thereby alleviating the computational bottleneck introduced by incorporating target-aware features. To achieve comprehensive multimodal integration, DMF combines user interest representations learned under the modality-centric and modality-enriched modeling strategies. Offline experiments on public and industrial datasets demonstrate the effectiveness of DMF. Moreover, DMF has been deployed on the product recommendation system of the international e-commerce platform Lazada, achieving relative improvements of 5.30% in CTCVR and 7.43% in GMV with negligible computational overhead.
format Preprint
id arxiv_https___arxiv_org_abs_2510_11066
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Decoupled Multimodal Fusion for User Interest Modeling in Click-Through Rate Prediction
Fan, Alin
Li, Hanqing
Lu, Sihan
Yuan, Jingsong
Zhang, Jiandong
Information Retrieval
Modern industrial recommendation systems improve recommendation performance by integrating multimodal representations from pre-trained models into ID-based Click-Through Rate (CTR) prediction frameworks. However, existing approaches typically adopt modality-centric modeling strategies that process ID-based and multimodal embeddings independently, failing to capture fine-grained interactions between content semantics and behavioral signals. In this paper, we propose Decoupled Multimodal Fusion (DMF), which introduces a modality-enriched modeling strategy to enable fine-grained interactions between ID-based collaborative representations and multimodal representations for user interest modeling. Specifically, we construct target-aware features to bridge the semantic gap across different embedding spaces and leverage them as side information to enhance the effectiveness of user interest modeling. Furthermore, we design an inference-optimized attention mechanism that decouples the computation of target-aware features and ID-based embeddings before the attention layer, thereby alleviating the computational bottleneck introduced by incorporating target-aware features. To achieve comprehensive multimodal integration, DMF combines user interest representations learned under the modality-centric and modality-enriched modeling strategies. Offline experiments on public and industrial datasets demonstrate the effectiveness of DMF. Moreover, DMF has been deployed on the product recommendation system of the international e-commerce platform Lazada, achieving relative improvements of 5.30% in CTCVR and 7.43% in GMV with negligible computational overhead.
title Decoupled Multimodal Fusion for User Interest Modeling in Click-Through Rate Prediction
topic Information Retrieval
url https://arxiv.org/abs/2510.11066