CaST-Bench: Benchmarking Causal Chain-Grounded Spatio-Temporal Reasoning for Video Question Answering
Quick Take
CaST-Bench introduces a novel benchmark for evaluating causal chain-grounded spatio-temporal reasoning in video question answering, featuring 2,066 questions across 1,015 videos. Current Vision-Language Models struggle with causal reasoning, highlighting the need for improved model capabilities in constructing grounded causal chains.
Key Points
- CaST-Bench consists of 2,066 questions over 1,015 videos with annotated causal chains.
- The benchmark focuses on identifying and localizing spatio-temporal evidence for causal reasoning.
- Current Vision-Language Models struggle with causal questions due to limited grounding capabilities.
- A comprehensive evaluation suite includes novel metrics for assessing visual evidence reasoning.
- Improving grounding can enhance model accuracy and user trust.
Article Content
From source RSS / original summaryarXiv:2605. 23216v1 Announce Type: new Abstract: 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.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →Evi-Steer: Learning to Steer Biomedical Vision-Language Models through Efficient and Generalizable Evidential Tuning
Evi-Steer introduces a novel evidential tuning framework for BiomedCLIP, achieving 0.11% parameter updates while enhancing uncertainty-aware fine-tuning. It outperforms state-of-the-art methods across 15 biomedical imaging datasets, proving effective in few-shot learning and domain shifts for clinical applications.