arXiv 2511.13909
Mind the Gap: Evaluating LLM Understanding of Human-Taught Road Safety Principles
By Chalamalasetti Kranti
Published 2025-11-17
Wiki summary
Explore the paper's summary, context, and related research on Papiers.
Following road safety norms is non-negotiable not only for humans but also for the AI systems that govern autonomous vehicles. In this work, we evaluate how well multi-modal large language models (LLMs) understand road safety concepts, specifically through schematic and illustrative representations. We curate a pilot dataset of images depicting traffic signs and road-safety norms sourced from school text books and u…