Benchmark RadarAI BENCHMARK PROFILE
MedBench
MedBench v5 evaluates clinical multimodal models across 63 tasks with process-oriented metrics including stressors and hallucination tracking.
- Released
- 2026-06-23
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
Addresses gaps in process visibility and hallucination detection in medical AI evaluation, offering a unified framework for capability profiling and stress testing.
Motivation
Existing medical AI benchmarks lack process visibility, atomic skill evaluation, and integrated hallucination detection.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.