COMET dissects the modality gap in CLAP embeddings using PLS-SVD, showing shared concept axes dominate similarity and proposing spectral truncation to mitigate the gap without training.
Quantifying structure in clip embeddings: A statistical framework for concept interpretation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
COMET: Concept Space Dissection of the Modality Gap in Audio-Text Multimodal Contrastive Embeddings
COMET dissects the modality gap in CLAP embeddings using PLS-SVD, showing shared concept axes dominate similarity and proposing spectral truncation to mitigate the gap without training.