CARE: Towards Clinical Accountability in Multi-Modal Medical Reasoning with an Evidence-Grounded Agentic Framework

Researchers introduce CARE, an evidence-grounded agentic framework for medical AI that improves clinical accountability by decomposing tasks into specialized modules rather than using black-box models. The system achieves 10.9% better accuracy than state-of-the-art models by incorporating explicit visual evidence and coordinated reasoning that mimics clinical workflows.

• →CARE framework addresses the black-box problem in medical AI by providing explicit, pixel-level evidence for diagnostic reasoning.
• →The system decomposes medical reasoning into specialized modules: entity proposal, visual grounding, and evidence-based reasoning.
• →CARE-Flow achieves 10.9% accuracy improvement over same-size state-of-the-art models on medical VQA benchmarks.
• →The framework uses reinforcement learning to align AI answers with supporting visual evidence.
• →An agentic coordinator provides dynamic planning and answer verification to ensure evidence-answer consistency.