视觉-语言因果推理中的抽象差距
The Abstraction Gap in Vision-Language Causal Reasoning
AI总结：
针对视觉-语言模型（VLM）生成因果解释时语言流畅性与忠实因果推理的混淆问题，提出双探针方法和抽象差距（AG）指标，通过CAGE基准评估发现多数模型存在显著AG，但通过预训练和架构选择可缩小差距。
摘要：
视觉-语言模型（VLM）能生成流畅的因果解释，但当前的评估无法区分语言合理性与忠实因果推理。我们提出一种双探针方法来分离这些属性。文本探针测量语言质量。链式文本探针要求模型首先生成显式因果链。抽象差距（AG）指标量化归一化的性能差异。在CAGE（因果抽象差距评估）基准上评估八个VLM，该基准包含跨越Pearl因果层次的5,500张图像上的49,500个问题，我们发现七个模型的AG超过0.50，文本得分为6-8，但链式得分低于2.5。在45,000个链式标注样本上进行微调未能缩小差距。然而，一个模型实现了接近零的AG。该能力存在于当前VLM架构中，并取决于预训练和架构选择。CAGE为评估VLM中的忠实因果推理提供了诊断工具。
英文摘要：
Vision-language models (VLMs) generate fluent causal explanations, but current evaluations cannot distinguish linguistic plausibility from faithful causal reasoning. We introduce a dual-probe methodology that isolates these properties. The Text-Only Probe measures linguistic quality. The Chain-Text Probe requires models to first generate explicit causal chains. The Abstraction Gap (AG) metric quantifies the normalized performance difference. Evaluating eight VLMs on CAGE (Causal Abstraction Gap Evaluation), a benchmark of 49,500 questions across 5,500 images spanning Pearl's causal hierarchy, we find seven models exhibit AG exceeding 0.50 with text scores of 6--8 but chain scores below 2.5. Fine-tuning on 45,000 chain-annotated examples fails to close the gap. However, one model achieves near-zero AG. The capability exists within current VLM architectures and depends on pretraining and architectural choices. CAGE provides a diagnostic tool for assessing faithful causal reasoning in VLMs.
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。
