Trust the Model: Compact VLMs as In-Context Judges for Image-Text Data Quality

VLMRe-assign community