ChartHal：一个细粒度评估大视觉语言模型在图表理解中幻觉的框架
ChartHal: A Fine-grained Framework Evaluating Hallucination of Large Vision Language Models in Chart Understanding
AI总结：
本文提出ChartHal基准，通过细粒度幻觉分类体系和1062个人工验证样本，系统评估大视觉语言模型在图表理解中的幻觉问题，并揭示现有顶尖模型准确率不足35%，亟需更稳健的缓解策略。
摘要：
大视觉语言模型（LVLMs）近年来取得了显著进展，但幻觉仍然是一个关键障碍，尤其是在图表理解中，这需要复杂的感知与认知能力以及严格的事实准确性。尽管先前的工作分别研究了幻觉和图表理解，但二者的交叉领域在很大程度上仍未得到探索。为填补这一空白，我们提出了ChartHal，这是一个基准，具有图表理解中幻觉场景的细粒度分类体系，以及一个经人工验证的包含1,062个样本的数据集。我们的评估表明，最先进的LVLMs在ChartHal上存在严重的幻觉，包括GPT-5和o4-mini等专有模型，其准确率分别仅为34.46%和22.79%。进一步分析揭示，涉及图表中缺失信息或与图表相矛盾信息的问题尤其容易触发幻觉，这凸显了迫切需要更稳健的缓解策略。代码和数据可在https://github.com/ymcui/ChartHal获取。
英文摘要：
Large Vision-Language Models (LVLMs) have recently demonstrated remarkable progress, yet hallucination remains a critical barrier, particularly in chart understanding, which requires sophisticated perceptual and cognitive abilities as well as rigorous factual accuracy. While prior work has investigated hallucinations and chart comprehension independently, their intersection remains largely unexplored. To address this gap, we present ChartHal, a benchmark that features a fine-grained taxonomy of hallucination scenarios in chart understanding, along with a human-validated dataset of 1,062 samples. Our evaluation shows that state-of-the-art LVLMs suffer from severe hallucinations on ChartHal, including proprietary models such as GPT-5 and o4-mini, which achieve only 34.46% and 22.79% accuracy, respectively. Further analysis reveals that questions involving information absent from or contradictory to charts are especially likely to trigger hallucinations, underscoring the urgent need for more robust mitigation strategies. Code and data are available at https://github.com/ymcui/ChartHal .
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
