Is Your VLM for Autonomous Driving Safety-Ready? A Comprehensive Benchmark for Evaluating External and In-Cabin Risks

TL;DR

This paper introduces DSBench, a comprehensive benchmark for evaluating vision-language models in autonomous driving safety, covering external and in-cabin risks, revealing performance gaps, and demonstrating improvements through fine-tuning.

Contribution

The paper presents the first unified benchmark for assessing VLM safety in autonomous driving, including a large dataset and evaluation toolkit to improve model safety performance.

Findings

VLMs show significant performance degradation in safety-critical scenarios.

Fine-tuning on the DSBench dataset improves VLM safety performance.

The benchmark covers 10 categories and 28 sub-categories of risks.

Abstract

Vision-Language Models (VLMs) show great promise for autonomous driving, but their suitability for safety-critical scenarios is largely unexplored, raising safety concerns. This issue arises from the lack of comprehensive benchmarks that assess both external environmental risks and in-cabin driving behavior safety simultaneously. To bridge this critical gap, we introduce DSBench, the first comprehensive Driving Safety Benchmark designed to assess a VLM's awareness of various safety risks in a unified manner. DSBench encompasses two major categories: external environmental risks and in-cabin driving behavior safety, divided into 10 key categories and a total of 28 sub-categories. This comprehensive evaluation covers a wide range of scenarios, ensuring a thorough assessment of VLMs' performance in safety-critical contexts. Extensive evaluations across various mainstream open-source and closed-source VLMs reveal significant performance degradation under complex safety-critical situations, highlighting urgent safety concerns. To address this, we constructed a large dataset of 98K instances focused on in-cabin and external safety scenarios, showing that fine-tuning on this dataset significantly enhances the safety performance of existing VLMs and paves the way for advancing autonomous driving technology. The benchmark toolkit, code, and model checkpoints will be publicly accessible.