VAUQ: Vision-Aware Uncertainty Quantification for LVLM Self-Evaluation
This addresses the limitation of existing self-evaluation methods for LVLMs, which rely on language priors and are ill-suited for vision-conditioned predictions, by providing a more reliable approach for safe deployment in real-world applications.
The paper tackles the problem of LVLM hallucination by proposing VAUQ, a vision-aware uncertainty quantification framework for self-evaluation that measures dependence on visual evidence, resulting in consistent outperformance over existing methods across multiple datasets.
Large Vision-Language Models (LVLMs) frequently hallucinate, limiting their safe deployment in real-world applications. Existing LLM self-evaluation methods rely on a model's ability to estimate the correctness of its own outputs, which can improve deployment reliability; however, they depend heavily on language priors and are therefore ill-suited for evaluating vision-conditioned predictions. We propose VAUQ, a vision-aware uncertainty quantification framework for LVLM self-evaluation that explicitly measures how strongly a model's output depends on visual evidence. VAUQ introduces the Image-Information Score (IS), which captures the reduction in predictive uncertainty attributable to visual input, and an unsupervised core-region masking strategy that amplifies the influence of salient regions. Combining predictive entropy with this core-masked IS yields a training-free scoring function that reliably reflects answer correctness. Comprehensive experiments show that VAUQ consistently outperforms existing self-evaluation methods across multiple datasets.