In-context learning for medical artificial intelligence.

In-context learning for medical AI is a promising but immature way for large language models to adapt to new tasks via examples in prompts, without changing parameters.

• Why it matters: It could enable rapid task switching and customization in clinical workflows, but performance is sensitive to prompt design and example selection, raising robustness, bias, and reproducibility concerns.
• What they did: This review synthesizes ICL applications across medical domains, including clinical question answering, information extraction, documentation, and multimodal imaging, and examines how approaches may evolve toward adaptive clinical systems.
• The result: It proposes a framework for safe clinical translation using risk-proportionate evaluation, governance, and continuous monitoring, supporting responsible deployment despite current limitations.

AI summary of the abstract · may contain errors · not medical advice · read the paper.