VideoQA in the Era of LLMs: An Empirical Study

TL;DR

This paper provides an empirical analysis of Video Large Language Models' performance in Video Question Answering, highlighting their strengths in content correlation and their weaknesses in temporal reasoning, robustness, and interpretability.

Contribution

It offers a comprehensive study of Video-LLMs' behavior in VideoQA, revealing their capabilities and limitations, and emphasizes the need for improved robustness and explainability.

Findings

Video-LLMs excel at correlating contextual cues and generating plausible answers.

Models struggle with reasoning about temporal content and grounding temporal moments.

They are insensitive to adversarial perturbations but sensitive to simple variations.

Abstract

Video Large Language Models (Video-LLMs) are flourishing and has advanced many video-language tasks. As a golden testbed, Video Question Answering (VideoQA) plays pivotal role in Video-LLM developing. This work conducts a timely and comprehensive study of Video-LLMs' behavior in VideoQA, aiming to elucidate their success and failure modes, and provide insights towards more human-like video understanding and question answering. Our analyses demonstrate that Video-LLMs excel in VideoQA; they can correlate contextual cues and generate plausible responses to questions about varied video contents. However, models falter in handling video temporality, both in reasoning about temporal content ordering and grounding QA-relevant temporal moments. Moreover, the models behave unintuitively - they are unresponsive to adversarial video perturbations while being sensitive to simple variations of candidate answers and questions. Also, they do not necessarily generalize better. The findings demonstrate Video-LLMs' QA capability in standard condition yet highlight their severe deficiency in robustness and interpretability, suggesting the urgent need on rationales in Video-LLM developing.