A regularizer derived from an information-bottleneck bound, essentially a mean-squared alignment loss, reduces modality-specific information and improves multimodal alignment and image captioning.
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
-
Aligning Multimodal Representations through an Information Bottleneck
A regularizer derived from an information-bottleneck bound, essentially a mean-squared alignment loss, reduces modality-specific information and improves multimodal alignment and image captioning.