Bridging the gap, not forcing the tie: dual-space alignment and fusion framework for toxic memes detection

Yulin Lei, Jin Yang, Huijia Liang, Tianrui Li

Abstract

Highlights•Geometry-area alignment preserves modality-specific semantics and structure.•Counterfactual weighting reduces bias from dominant modalities.•Adaptive multi-task fusion integrates semantic, alignment, and decision cues.•ALFUS shows strong cross-lingual performance on toxic meme datasets.
BibTeX
@article{lei2026bridging,
  author = {Yulin Lei and Jin Yang and Huijia Liang and Tianrui Li},
  title = {{Bridging the gap, not forcing the tie: dual-space alignment and fusion framework for toxic memes detection}},
  journal = {Information Fusion},
  year = {2026},
  url = {https://openreview.net/forum?id=OG0yWJHL6i},
}

Loading related content...

Figure