Despite the remarkable multimodal capabilities of Large Vision-Language Models (LVLMs), discrepancies often occur between visual inputs and textual outputs--a phenomenon we term visual hallucination. This critical reliability gap poses substantial risks in safety-critical Artificial Intelligence (AI) applications, necessitating a comprehensive evaluation benchmark and effective detection methods. Firstly, we observe that existing visual-centric hallucination benchmarks mainly assess LVLMs from a perception perspective, overlooking hallucinations arising from advanced reasoning capabilities. We develop the Perception-Reasoning Evaluation Hallucination (PRE-HAL) dataset, which enables the systematic evaluation of both perception and reasoning capabilities of LVLMs across multiple visual semantics, such as instances, scenes, and relations. Comprehensive evaluation with this new benchmark exposed more visual vulnerabilities, particularly in the more challenging task of relation reasoning. To address this issue, we propose, to the best of our knowledge, the first Dempster-Shafer theory (DST)-based visual hallucination detection method for LVLMs through uncertainty estimation. This method aims to efficiently capture the degree of conflict in high-level features at the model inference phase. Specifically, our approach employs simple mass functions to mitigate the computational complexity of evidence combination on power sets. We conduct an extensive evaluation of state-of-the-art LVLMs, LLaVA-v1.5, mPLUG-Owl2 and mPLUG-Owl3, with the new PRE-HAL benchmark. Experimental results indicate that our method outperforms five baseline uncertainty metrics, achieving average AUROC improvements of 4%, 10%, and 7% across three LVLMs. Our code is available at https://github.com/HT86159/Evidential-Conflict.

翻译：尽管大型视觉语言模型（LVLMs）展现出卓越的多模态能力，但视觉输入与文本输出之间常存在不一致——我们将此现象称为视觉幻觉。这一关键的可信度差距在安全关键的人工智能（AI）应用中构成重大风险，亟需建立全面的评估基准和有效的检测方法。首先，我们观察到现有的以视觉为中心的幻觉基准主要从感知角度评估LVLMs，忽视了由高级推理能力引发的幻觉。我们开发了感知-推理评估幻觉（PRE-HAL）数据集，该数据集能够系统评估LVLMs在多种视觉语义（如实例、场景和关系）上的感知与推理能力。使用这一新基准进行的全面评估揭示了更多的视觉脆弱性，尤其是在更具挑战性的关系推理任务中。为解决此问题，据我们所知，我们首次提出了一种基于Dempster-Shafer理论（DST）的、通过不确定性估计进行LVLMs视觉幻觉检测的方法。该方法旨在模型推理阶段高效捕获高层特征中的冲突程度。具体而言，我们的方法采用简单的质量函数来降低幂集上证据组合的计算复杂度。我们使用新的PRE-HAL基准对最先进的LVLMs（LLaVA-v1.5、mPLUG-Owl2和mPLUG-Owl3）进行了广泛评估。实验结果表明，我们的方法优于五种基线不确定性度量，在三种LVLMs上平均AUROC分别提升了4%、10%和7%。我们的代码可在 https://github.com/HT86159/Evidential-Conflict 获取。