RXL-RADSet is a comprehensive benchmark dataset designed for evaluating language models on radiology report understanding and generation tasks.
Radiologist-Validated: All reports have been validated by experienced radiologists
Synthetic Generation: Reports are generated using controlled synthetic methods ensuring quality and diversity
Multi-Modal Coverage: Includes reports across CT, MRI, Mammography, and Ultrasound
10 RADS Systems: Covers NI-RADS, TI-RADS, Lung-RADS… See the full description on the dataset page:
https://huggingface.co/datasets/RadioX-Labs/RADSet.
