Empirical benchmarking finds VLMs for medical image quality assessment are highly sensitive to pixelation corruption and textual attributes like institutional prestige, indicating limited reliability and objectivity.
arXiv preprint arXiv:250607044
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Assessing VLM Reliability for Medical Image Quality Evaluation Under Corruption and Bias
Empirical benchmarking finds VLMs for medical image quality assessment are highly sensitive to pixelation corruption and textual attributes like institutional prestige, indicating limited reliability and objectivity.