MVCL-DAF++ combines prototype-based contrastive learning with coarse-to-fine attention fusion and reports improved accuracy and weighted F1 on MIntRec and MIntRec2.0.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MVCL-DAF++: Enhancing Multimodal Intent Recognition via Prototype-Aware Contrastive Alignment and Coarse-to-Fine Dynamic Attention Fusion
MVCL-DAF++ combines prototype-based contrastive learning with coarse-to-fine attention fusion and reports improved accuracy and weighted F1 on MIntRec and MIntRec2.0.