The original title is a survey title, not a news headline. I need to rewrite it as a news headline for a mobile feed.

Short summary

This survey examines how LLMs can support clinical reasoning using a five-level competency framework. Benchmarking 18 SOTA models reveals specialist medical models excel at diagnosis while general models lead in decision support. Key challenges—hallucination, data limitations, grounding—require further work for clinically safe, workflow-ready systems.

• •Survey introduces Miller's Pyramid-based competency framework for medical reasoning with LLMs
• •Benchmarks 18 SOTA models: specialist medical models win on diagnosis, general models excel at decision support
• •Identifies critical challenges—hallucination, data gaps, grounding issues—blocking production clinical deployment

Generated with AI, which can make mistakes.