Vision-language text encoders encode less syntactic structure than text-only encoders, and contrastive pre-training rather than model size or data volume accounts for most of the deficit.
On the difference of bert-style and clip-style text encoders
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Seeing Syntax: Uncovering Syntactic Learning Limitations in Vision-Language Models
Vision-language text encoders encode less syntactic structure than text-only encoders, and contrastive pre-training rather than model size or data volume accounts for most of the deficit.