CaST-Bench：面向视频问答的因果链视觉定位时空推理基准
CaST-Bench: Benchmarking Causal Chain-Grounded Spatio-Temporal Reasoning for Video Question Answering
该基准要求模型在视频中定位因果链的时空证据（时序片段和边界框轨迹），为视觉语言模型的可解释推理和减少虚假关联提供了可评估的测试平台，其标注管线和评估指标可直接用于训练和评测更透明的视频问答模型。
原文依据 CaST-Bench presents complex causal questions that require models to identify and localize a chain of multiple spatio-temporal evidences.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频中的因果推理对视觉语言模型（VLM）而言是一项重大挑战，因为它要求超越表面感知，深入理解因果机制。然而，现有基准很少提供严格评估这一能力所需的细粒度、有视觉定位的证据。为填补这一空白，我们提出了 CaST-Bench，一个用于因果链视觉定位时空视频推理的基准。CaST-Bench 提出了复杂的因果问题，要求模型识别并定位由多个时空证据组成的链。通过人机协同流程，我们构建了一个高质量数据集，包含 1,015 个视频上的 2,066 个问题，因果链以时序片段和边界框轨迹进行标注。此外，我们设计了一套全面的评估套件，采用新颖的指标，不仅评估答案正确性，还评估基于视觉证据的推理能力。这种视觉定位对于通过减轻虚假关联来提高准确性，以及通过使模型更透明来增强用户信任至关重要。我们的实验表明，当前 VLM 在因果问题上表现不佳，主要归因于它们构建精确且基于视觉定位的因果链的能力有限。这凸显了改进未来 VLM 的一个重要方向。
英文原文摘要 1322 字符
Cause-and-effect reasoning in video is a significant challenge for Vision-Language Models (VLMs), as it requires going beyond surface-level perception to a deeper understanding of causal mechanisms. However, existing benchmarks rarely provide the fine-grained, grounded evidence needed to rigorously evaluate this capability. To address this gap, we introduce CaST-Bench, a benchmark for Causal Chain-Grounded Spatio-Temporal Video Reasoning. CaST-Bench presents complex causal questions that require models to identify and localize a chain of multiple spatio-temporal evidences. Through a human-AI collaborative pipeline, we construct a high-quality dataset of 2,066 questions over 1,015 videos, with causal chains annotated by temporal segments and bounding-box tracks. Furthermore, we design a comprehensive evaluation suite with novel metrics that assess not only answer correctness but also the capability for visual evidence grounded reasoning. This grounding is crucial for improving accuracy by mitigating spurious correlations and for enhancing user trust by making models more transparent. Our experiments show that current VLMs struggle with causal questions, largely due to their limited ability to construct precise and grounded causal chains. This highlights an important direction for improving future VLMs.