视觉-语言模型对连续驾驶场景的理解有多好？一项敏感性研究
How Well Do Vision-Language Models Understand Sequential Driving Scenes? A Sensitivity Study
AI总结：
本文通过系统分析视觉-语言模型（VLMs）在连续驾驶场景中的表现，揭示了输入配置对模型性能的影响，发现即使顶级模型在连续驾驶场景中的准确率仅为57%，远低于人类在相似约束下的65%，并暴露了VLMs在理解车辆动态和时间关系上的显著差距。
摘要：
视觉-语言模型（VLMs）越来越多地被提出用于自动驾驶任务，但它们在连续驾驶场景中的性能仍然缺乏充分的描述，尤其是在输入配置如何影响其能力方面。我们介绍了VENUSS（VLM评估在理解连续场景上），这是一个用于系统敏感性分析VLM在连续驾驶场景中性能的框架，为未来研究建立了基准。基于现有数据集，VENUSS从驾驶视频中提取时间序列，并在自定义类别中生成结构化评估。通过在2,600多个场景上比较25多个现有VLMs，我们揭示了即使顶级模型也只能达到57%的准确率，这并未达到在相似约束下人类的65%表现，并暴露了显著的能力差距。我们的分析表明，VLMs在静态物体检测方面表现优异，但在理解车辆动态和时间关系方面则存在困难。VENUSS提供了首次系统性的VLM敏感性分析，专注于输入图像配置（如分辨率、帧数、时间间隔、空间布局和展示模式）如何影响连续驾驶场景中的性能。补充材料可在https://TUM-AVS.github.io/VENUSS/上获得。
英文摘要：
Vision-Language Models (VLMs) are increasingly proposed for autonomous driving tasks, yet their performance on sequential driving scenes remains poorly characterized, particularly regarding how input configurations affect their capabilities. We introduce VENUSS (VLM Evaluation oN Understanding Sequential Scenes), a framework for systematic sensitivity analysis of VLM performance on sequential driving scenes, establishing baselines for future research. Building upon existing datasets, VENUSS extracts temporal sequences from driving videos, and generates structured evaluations across custom categories. By comparing 25+ existing VLMs across 2,600+ scenarios, we reveal how even top models achieve only 57% accuracy, not matching human performance under similar constraints (65%) and exposing significant capability gaps. Our analysis shows that VLMs excel with static object detection but struggle with understanding vehicle dynamics and temporal relations. VENUSS offers the first systematic sensitivity analysis of VLMs focused on how input image configurations - resolution, frame count, temporal intervals, spatial layouts, and presentation modes - affect performance on sequential driving scenes. Supplementary material available at https://TUM-AVS.github.io/VENUSS/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
