From Theory to Decision Rule: Calibrating the Noisy-Label Crossover for Vision-Language Model Weak Supervision Across Three Medical-Imaging Benchmarks

In medical imaging, small sets of expert labels are often combined with zero-shot pseudo-labels from a vision-language model (VLM). When the expert-labeled set is very small, the extra pseudo-labels reduce variance and improve accuracy. Once a classifier trained on expert labels alone reaches the ac