语气至关重要：语言语气对VLMs幻觉影响的研究
Tone Matters: The Impact of Linguistic Tone on Hallucination in VLMs
AI总结：
本文研究了提示语气对VLMs幻觉的影响，通过Ghost-100数据集发现幻觉率与提示强度非线性相关，揭示模型在处理结构性强制时的局限性。
摘要：
视觉-语言模型（VLMs）正被越来越多地用于需要可靠视觉基础的安全关键应用。然而，这些模型常常会产生图像中没有的细节来满足用户提示。尽管最近的数据集和基准已被引入以评估VLMs中的系统性幻觉，但许多幻觉行为仍然未被充分描述。特别是，先前的工作主要集中在对象存在与否上，不清楚提示语的表达方式和结构约束如何系统性地引起幻觉。在本文中，我们研究了不同形式的提示压力如何影响幻觉行为。我们引入了Ghost-100，一个程序生成的合成场景数据集，在其中关键的视觉细节被有意删除，从而能够对基于缺席的幻觉进行受控分析。使用一个结构化的5级提示强度框架，我们从中性查询到有毒需求和刚性格式约束的变化。我们评估了三个代表性的开源VLMs：MiniCPM-V 2.6-8B、Qwen2-VL-7B和Qwen3-VL-8B。在所有三个模型中，幻觉率并不随提示强度单调增加。所有模型在更高强度水平上都表现出减少，但在不同阈值下，尽管并非所有模型在最大压力下都持续减少。这些结果表明，当前的安全对齐更有效于检测语义敌意而非结构性强制，揭示了模型在处理合规压力时的特定限制。我们的数据集可在：https://github.com/bli1/tone-matters上获得。
英文摘要：
Vision-Language Models (VLMs) are increasingly used in safety-critical applications that require reliable visual grounding. However, these models often hallucinate details that are not present in the image to satisfy user prompts. While recent datasets and benchmarks have been introduced to evaluate systematic hallucinations in VLMs, many hallucination behaviors remain insufficiently characterized. In particular, prior work primarily focuses on object presence or absence, leaving it unclear how prompt phrasing and structural constraints can systematically induce hallucinations. In this paper, we investigate how different forms of prompt pressure influence hallucination behavior. We introduce Ghost-100, a procedurally generated dataset of synthetic scenes in which key visual details are deliberately removed, enabling controlled analysis of absence-based hallucinations. Using a structured 5-Level Prompt Intensity Framework, we vary prompts from neutral queries to toxic demands and rigid formatting constraints. We evaluate three representative open-weight VLMs: MiniCPM-V 2.6-8B, Qwen2-VL-7B, and Qwen3-VL-8B. Across all three models, hallucination rates do not increase monotonically with prompt intensity. All models exhibit reductions at higher intensity levels at different thresholds, though not all show sustained reduction under maximum coercion. These results suggest that current safety alignment is more effective at detecting semantic hostility than structural coercion, revealing model-specific limitations in handling compliance pressure. Our dataset is available at: https://github.com/bli1/tone-matters
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。
