ACL Findings2026
Not All Modalities at Once: Dynamic Dropout and Bidirectional Fusion for Robust Multi-modal Knowledge Graph Completion
Jiashun Peng, Fu Zhang, Hongzhi Chen, Jingwei Cheng, Yingsong Ning, Xiaoke Wang
Abstract
Multi-modal Knowledge Graph Completion (MKGC) aims to infer missing links in multimodal knowledge graphs by leveraging structured triples together with auxiliary modalities such as text and images. Existing MKGC methods typically train with all modalities available, implicitly assuming consistent complementarity; however, this practice often induces modality dependence and modality competition under heterogeneous noise, which can hinder robust multi-modal fusion and limit overall performance. To address these issues, we propose MDBGF, a Modality Dropout and Bidirectional Gated Fusion framework for MKGC. MDBGF introduces a dynamic, probability-based modality dropout schedule. When the dropout is activated, MDBGF drops either the textual or visual modality during training while always preserving the structural information, encouraging the model to reduce over-reliance on any single auxiliary modality and to learn complementary cues under missing-modality conditions. When the dropout is not activated (i.e., all modalities are present), we further design a bidirectional gated fusion mechanism that enables mutual modulation between textual and visual modalities, enhancing cross-modal interaction and flexible fusion. In addition, we propose an adaptive proportional hybrid negative sampling strategy to strengthen MDBGF's discriminative ability on hard negatives. Experiments on three benchmarks show that MD-BGF consistently outperforms existing baselines and achieves new state-of-the-art results. Our code is available at https://github. com/ferryman-ship/MDBGF.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- VISTA: Visual-Textual Knowledge Graph Representation Learning cites IMF: Interactive Multimodal Fusion Model for Link Prediction
- VISTA: Visual-Textual Knowledge Graph Representation Learning cites OTKGE: Multi-modal Knowledge Graph Embeddings via Optimal Transport
Builds on8
- Modality Competition: What Makes Joint Training of Multi-modal Network Fail in Deep Learning? (Provably)Yu Huang, Junyang Lin, Chang Zhou, Hongxia Yang et al.ICML 2022 · 168 citations
- OTKGE: Multi-modal Knowledge Graph Embeddings via Optimal TransportZongsheng Cao, Qianqian Xu, Zhiyong Yang, Yuan He et al.NeurIPS 2022 · 117 citations
- IMF: Interactive Multimodal Fusion Model for Link PredictionXinhang Li, Xiangyu Zhao, Jiaxing Xu, Yong Zhang et al.WWW 2023 · 113 citations
- Relation-enhanced Negative Sampling for Multimodal Knowledge Graph CompletionDerong Xu, Tong Xu, Shiwei Wu, Jingbo Zhou et al.ACM MM 2022 · 89 citations
- VISTA: Visual-Textual Knowledge Graph Representation LearningJaejun Lee, Chanyoung Chung, Hochang Lee, Sungho Jo et al.EMNLP Findings 2023 · 58 citations
Related papers
- NativE: Multi-modal Knowledge Graph Completion in the WildYichi Zhang, Zhuo Chen, Lingbing Guo, Yajing Xu et al.SIGIR 2024 · 39 citations
- HFR-MKGC: Hierarchical Fusion Reasoning with MLLMs for Multi-modal Knowledge Graph CompletionDi Wang, Junping Du, Zhe Xue, Meiyu Liang et al.AAAI 2026
- MMKGR: Multi-hop Multi-modal Knowledge Graph ReasoningShangfei Zheng, Weiqing Wang, Jianfeng Qu, Hongzhi Yin et al.ICDE 2023 · 40 citations
- Collaboration of Fusion and Independence: Hypercomplex-driven Robust Multi-Modal Knowledge Graph CompletionZhiqiang Liu, Yichi Zhang, Mengshu Sun, Lei Liang et al.ACL 2026 · 1 citation
- Mixed-Curvature Multi-Modal Knowledge Graph CompletionYuxiao Gao, Fuwei Zhang, Zhao Zhang, Xiaoshuang Min et al.AAAI 2025 · 5 citations
