arxivcs.CV2026-06-25
CORTEX: A Structured Reasoning Benchmark for Trustworthy 3D Chest CT MLLMs
Hashmat Shadab Malik, Anees Ur Rehman Hashmi, Numan Saeed, Muzammal Naseer, Salman Khan, Christoph Lippert
Reasoning in multimodal large language models (MLLMs) has shown strong promise in medical imaging. However, this reasoning is usually free-form text judged only by its final answer, making it hard to interpret and verify, especially in 3D radiology, where a diagnosis should be tr…