Radiology's Last Exam 2.0 (RadLE-C)
Multimodal
Radiology's Last Exam 2.0 (RadLE 2.0) evaluates autonomous single-image spot diagnosis across 200 radiology cases. Its primary RadLE-C Confidence Weighted Index rewards correct diagnoses and penalizes incorrect diagnoses in proportion to the model's reported confidence, with an explicit 'I don't know' receiving no credit or penalty.