Bench2ADVLM: A Closed-Loop Benchmark for Vision-language Models in Autonomous Driving

TL;DR

Bench2ADVLM is a comprehensive closed-loop evaluation framework for vision-language models in autonomous driving, enabling real-time testing in simulation and physical environments to better assess safety and performance.

Contribution

It introduces a hierarchical closed-loop evaluation pipeline and a physical testing platform for ADVLMs, addressing the gap in current open-loop assessment methods.

Findings

Existing ADVLMs show limited performance in closed-loop settings.

The framework uncovers potential failure modes in real-world scenarios.

Physical testing validates simulation-based evaluations.

Abstract

Vision-Language Models (VLMs) have recently emerged as a promising paradigm in autonomous driving (AD). However, current performance evaluation protocols for VLM-based AD systems (ADVLMs) are predominantly confined to open-loop settings with static inputs, neglecting the more realistic and informative closed-loop setting that captures interactive behavior, feedback resilience, and real-world safety. To address this, we introduce Bench2ADVLM, a unified hierarchical closed-loop evaluation framework for real-time, interactive assessment of ADVLMs across both simulation and physical platforms. Inspired by dual-process theories of cognition, we first adapt diverse ADVLMs to simulation environments via a dual-system adaptation architecture. In this design, heterogeneous high-level driving commands generated by target ADVLMs (fast system) are interpreted by a general-purpose VLM (slow system) into standardized mid-level control actions suitable for execution in simulation. To bridge the gap between simulation and reality, we design a physical control abstraction layer that translates these mid-level actions into low-level actuation signals, enabling, for the first time, closed-loop testing of ADVLMs on physical vehicles. To enable more comprehensive evaluation, Bench2ADVLM introduces a self-reflective scenario generation module that automatically explores model behavior and uncovers potential failure modes for safety-critical scenario generation. Overall, Bench2ADVLM establishes a hierarchical evaluation pipeline that seamlessly integrates high-level abstract reasoning, mid-level simulation actions, and low-level real-world execution. Experiments on diverse scenarios across multiple state-of-the-art ADVLMs and physical platforms validate the diagnostic strength of our framework, revealing that existing ADVLMs still exhibit limited performance under closed-loop conditions.