AI RESEARCH
RADAR: A Multimodal Benchmark for 3D Image-Based Radiology Report Review
arXiv CS.CV
•
ArXi:2603.06681v1 Announce Type: new Radiology reports for the same patient examination may contain clinically meaningful discrepancies arising from interpretation differences, reporting variability, or evolving assessments. Systematic analysis of such discrepancies is important for quality assurance, clinical decision, and multimodal model development, yet remains limited by the lack of standardized benchmarks.