A multimodal relation extraction system that models text and image features as Gaussian distributions, denoises them with uncertainty-aware contrastive learning, and aligns the distributions with symmetric KL achieves state-of-the-art F1 on UMRE, MORE, and MNRE.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
UG-UMRE: Uncertainty-Guided Modality Augmentation and Distributional Calibration for Unified Multimodal Relation Extraction
A multimodal relation extraction system that models text and image features as Gaussian distributions, denoises them with uncertainty-aware contrastive learning, and aligns the distributions with symmetric KL achieves state-of-the-art F1 on UMRE, MORE, and MNRE.