Abstract: Large vision-language models (LVLMs) have demonstrated remarkable capabilities in autonomous driving scene understanding tasks. However, these models occasionally generate hallucinatory ...