Detecting Clinical Hallucinations in LVLMs via Counterfactual Visual Grounding Uncertainty

2026-06-26

AI Summary

This work addresses the susceptibility of large vision-language models (LVLMs) to generating hallucinations—clinically implausible statements inconsistent with input medical images. The authors propose a model-agnostic, non-intrusive hallucination detection method that requires no access to internal model representations or architectural modifications. Their approach employs a medical-adapted visual grounding verifier to localize clinical entities mentioned in generated text and introduces counterfactual perturbations to quantify uncertainty in visual evidence. By integrating entity-level confidence scores with localization overlap analysis, the method enables interpretable and cross-model hallucination identification. Experiments demonstrate that it significantly outperforms existing baselines across diverse medical imaging modalities and LVLM architectures, exhibiting strong generalization and providing traceable visual grounding evidence for detected hallucinations.

Abstract

Large vision-language models (LVLMs) are increasingly used for clinical image understanding, yet they remain vulnerable to \emph{hallucinations}--producing textual findings or attributes not supported by the image. We present a vision-traceable hallucination detection framework that audits arbitrary LVLM responses via visual evidence grounding, requiring neither modification nor internal access to the hidden states of LVLMs. Given an LVLM response, we extract visually verifiable entities and use a medical-domain-adapted Qwen-VL grounding verifier to localize each entity on the input image. To enhance the robustness of our detection method, we introduce a counterfactual entity perturbation method and estimate visual evidence uncertainty by contrasting factual and counterfactual grounding results. Specifically, we compute an entity-level uncertainty score from the positive confidence, counterfactual confidence, and their grounding overlap for binary hallucination decision-making. Experiments on multiple medical imaging modalities and LVLM backbones demonstrate that our method consistently improves hallucination detection performance over recent baselines, while providing interpretable localization evidence and strong cross-model transferability. Code and dataset are available at https://github.com/Agentic-CliniAI/CounterVHD.

Problem

• clinical hallucinations
• large vision-language models
• visual grounding
• medical image understanding
• hallucination detection

Innovation

• counterfactual visual grounding
• hallucination detection
• visual uncertainty estimation
• medical vision-language models
• entity-level verification