Benchmark RadarMEDLAYXPLAIN
MedLayXPlain is a benchmark for medical lay language generation, pairing medical images with expert and lay captions across 122,789 samples from 8 imaging modalities. It introduces a 3B evaluator model scoring expert-lay alignment on five attributes.
- Released
- 2026-06-19
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
Addresses the gap between expert-level medical image descriptions and patient-accessible language, crucial for patient education and shared decision-making under recent regulations. Provides a standardized evaluation for medical VLMs in patient-facing communication.
Motivation
Medical Vision-Language Models (Med-VLMs) achieve strong expert-level performance, yet their ability to generate patient-accessible descriptions remains underexplored.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.