VirtueBench: Evaluating Trustworthiness under Uncertainty in Long Video Understanding

TL;DR

VirtueBench is a new benchmark designed to evaluate the trustworthiness of vision-language models in long video understanding, focusing on their ability to refuse answering when uncertain, thus promoting more reliable AI systems.

Contribution

The paper introduces VirtueBench, a benchmark that assesses model trustworthiness under uncertainty by constructing multiple frame-sampling levels and providing ground truths for answerability.

Findings

Refusal accuracy varies significantly across models.

Most models tend to answer even when uncertain.

Benchmark reveals models' refusal behaviors and reliability issues.

Abstract

Recent Vision-Language Models (VLMs) have made remarkable progress in multimodal understanding tasks, yet their evaluation on long video understanding remains unreliable. Due to limited frame inputs, key frames necessary for answering the question may be missing from the model's input. However, models that truthfully refuse to answer under such uncertainty are marked as incorrect, while those that guess may coincidentally produce the correct answer and thus obtain deceptively higher accuracy, leading to misleading evaluation results and encouraging models to guess rather than respond honestly. To address this issue, we introduce VirtueBench, a benchmark explicitly designed to assess model trustworthiness under uncertainty. VirtueBench constructs multiple frame-sampling levels for each video and provides ground truths that distinguish between answerable and unanswerable cases. Evaluations on 25 open-source and commercial VLMs reveal distinct refusal behaviors across different model families, with refusal accuracy ranging from over 70% in the best models to nearly 0% in the worst. Moreover, most models exhibit a substantial drop in refusal when the prompt does not explicitly require them to do so. These findings highlight the need for developing trustworthy VLMs for multimodal understanding, guided by benchmarks and leaderboards that emphasize reliability and trustworthiness.