.png){kind=icon}
.png){kind=icon}
Return
Curriculum-meta learning for unbiased multimodal relation extraction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jvcir.2026.104834.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We propose CMRE for long-tail multimodal relation extraction. • A semantic-guided curriculum strategy enables easy-to-hard relation learning. • Meta-learning and tail adaptation improve low-resource tail relation recognition. • MLLM-based augmentation enriches tail-class representations. • CMRE achieves up to 4.6% average Macro F1 improvement on two benchmarks.
Keywords:
CMRE
long-tail multimodal relation extraction
semantic-guided curriculum strategy
meta-learning
tail adaptation
Journal
IF:
3.1
Papers:
414
Citations:
5.6K