Are Large Vision Language Models Truly Grounded in Medical Images? Evidence from Italian Clinical Visual Question Answering
This work addresses the reliability of VLMs for medical applications, highlighting critical differences in model robustness that are important for clinical deployment, though it is incremental as it focuses on evaluating existing models.
The study investigated whether large vision language models (VLMs) are genuinely grounded in medical images by testing four state-of-the-art models on Italian clinical visual question answering, revealing that GPT-4o showed the strongest visual grounding with a 27.9 percentage point accuracy drop when images were removed, while others had modest drops.
Large vision language models (VLMs) have achieved impressive performance on medical visual question answering benchmarks, yet their reliance on visual information remains unclear. We investigate whether frontier VLMs demonstrate genuine visual grounding when answering Italian medical questions by testing four state-of-the-art models: Claude Sonnet 4.5, GPT-4o, GPT-5-mini, and Gemini 2.0 flash exp. Using 60 questions from the EuropeMedQA Italian dataset that explicitly require image interpretation, we substitute correct medical images with blank placeholders to test whether models truly integrate visual and textual information. Our results reveal striking variability in visual dependency: GPT-4o shows the strongest visual grounding with a 27.9pp accuracy drop (83.2% [74.6%, 91.7%] to 55.3% [44.1%, 66.6%]), while GPT-5-mini, Gemini, and Claude maintain high accuracy with modest drops of 8.5pp, 2.4pp, and 5.6pp respectively. Analysis of model-generated reasoning reveals confident explanations for fabricated visual interpretations across all models, suggesting varying degrees of reliance on textual shortcuts versus genuine visual analysis. These findings highlight critical differences in model robustness and the need for rigorous evaluation before clinical deployment.