VLMs perform far worse than their own visual encoders on vision-centric tasks because the language model fails to use accessible visual information and instead follows its language priors.
Hpatches: A benchmark and evaluation of handcrafted and learned local descriptors
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Hidden in plain sight: VLMs overlook their visual representations
VLMs perform far worse than their own visual encoders on vision-centric tasks because the language model fails to use accessible visual information and instead follows its language priors.