Imbalanced multimodal learning that prioritizes the performance-dominant modality via unimodal ranking and asymmetric gradient modulation outperforms balanced approaches.
Facilitating multimodal classification via dynamically learning modality gap,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
PDMP: Rethinking Balanced Multimodal Learning via Performance-Dominant Modality Prioritization
Imbalanced multimodal learning that prioritizes the performance-dominant modality via unimodal ranking and asymmetric gradient modulation outperforms balanced approaches.