Med-StepBench: A Hierarchical Reasoning Framework for Evaluating Hallucinations in Medical Vision-Language Models
2026-05-11 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created Med-StepBench, a large test set to check how well vision-language models (VLMs) understand medical images, especially 3D cancer scans. They found that while these models often give correct overall diagnoses, they can make mistakes in the detailed reasoning steps, which might be risky in real medical settings. Their benchmark breaks down the diagnosis process into stages to catch these hidden errors. They also discovered that the models get tricked by false but believable explanations, leading to more mistakes. Overall, the authors show that current models have trouble reliably linking image details to step-by-step clinical reasoning.
vision-language modelsmedical image understandinghallucination detection3D oncological PET/CTclinical reasoningbenchmarkadversarial examplesdiagnostic stagesmedical AI safetyvolumetric imaging
Authors
Minh Khoi Nguyen, Dai Lam Le, Amir Reza Jafari, Tuan Dung Nguyen, Mai Hong Son, Mai Huy Thong, Quang Huy Nguyen, Thanh Trung Nguyen, Reza Farahbakhsh, Noel Crespi, Phi Le Nguyen
Abstract
Large vision-language models (VLMs) demonstrate strong performance in medical image understanding, but frequently generate clinically plausible yet incorrect statements, raising significant safety concerns. Existing medical hallucination benchmarks primarily focus on 2D imaging with one-shot diagnostic questions, offering limited insight into whether predictions are grounded in correct localization and abnormality identification, allowing critical reasoning errors to remain hidden behind seemingly correct diagnoses. We introduce Med-StepBench, the first large-scale benchmark for step-wise hallucination detection in 3D oncological PET/CT, comprising over 12,000 images and more than 1,000,000 image-statement pairs across volumetric and multi-view 2D data, which decomposes clinical reasoning into four expert-designed diagnostic stages. Using clinician-verified annotations, we perform the first step-level evaluation of general-purpose and medical VLMs, revealing systematic failure modes obscured by aggregate accuracy metrics. Furthermore, we show that current VLMs are highly susceptible to adversarial yet clinically plausible intermediate explanations, which significantly amplify hallucinations despite contradictory visual evidence. Together, our findings highlight fundamental limitations in grounding multi-step clinical reasoning and establish Med-StepBench as a rigorous benchmark for developing safer and more reliable medical VLMs.