Benchmark RadarAI BENCHMARK PROFILE
DDX-TRACE
DDX-TRACE evaluates medical diagnostic trajectories in multimodal neuroradiology over 211 physician-adjudicated cases, where models request imaging studies, update differential diagnoses, and stop with a localized final diagnosis under hidden evidence.
- Released
- 2026-05-22
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
Traditional medical AI benchmarks reward final answers, overlooking workup quality. DDX-TRACE measures evidence-supported diagnostic processes, providing a more realistic and actionable evaluation for clinical decision support.
Motivation
Medical diagnosis is not a single prediction from a fully specified vignette.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.