Trust the Model: Compact VLMs as In-Context Judges for Image-Text Data Quality Paper • 2507.20156 • Published Jul 27