Towards Video Thinking Test: A Holistic Benchmark for Advanced Video Reasoning and Understanding

TL;DR

This paper introduces the Video Thinking Test (Video-TT), a comprehensive benchmark with real-world videos and adversarial questions to evaluate the reasoning and robustness of video large language models compared to human understanding.

Contribution

The paper presents Video-TT, a new holistic benchmark for assessing advanced video reasoning and robustness in video LLMs, highlighting existing gaps in performance.

Findings

Video-LLMs lag behind humans in understanding complex videos.

Video-TT effectively reveals robustness issues in current models.

Benchmark includes 1,000 YouTube Shorts with diverse questions.

Abstract

Human intelligence requires correctness and robustness, with the former being foundational for the latter. In video understanding, correctness ensures the accurate interpretation of visual content, and robustness maintains consistent performance in challenging conditions. Despite advances in video large language models (video LLMs), existing benchmarks inadequately reflect the gap between these models and human intelligence in maintaining correctness and robustness in video interpretation. We introduce the Video Thinking Test (Video-TT), to assess if video LLMs can interpret real-world videos as effectively as humans. Video-TT reflects genuine gaps in understanding complex visual narratives, and evaluates robustness against natural adversarial questions. Video-TT comprises 1,000 YouTube Shorts videos, each with one open-ended question and four adversarial questions that probe visual and narrative complexity. Our evaluation shows a significant gap between video LLMs and human performance.