A new 27-dataset benchmark shows current vision-language models average about 50% on high-resolution image tasks and display uneven regional attention in synthetic grid tests.
Ef- ficient high-resolution deep learning: A survey
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
HRScene: How Far Are VLMs from Effective High-Resolution Image Understanding?
A new 27-dataset benchmark shows current vision-language models average about 50% on high-resolution image tasks and display uneven regional attention in synthetic grid tests.