BD-HazardVLM：评估视觉语言模型在孟加拉国道路场景中的潜在防御性驾驶危险
原标题：BD-HazardVLM: Probing Vision-Language Models for Latent Defensive-Driving Hazards in Bangladesh Road Scenes
AI 摘要
研究者提出 BD-HazardVLM，一个包含 500 张孟加拉国道路场景图像的诊断基准，用于评估视觉语言模型的防御性驾驶推理能力。他们用统一零样本任务定义和结构化输出模式测试了 Qwen2.5-VL-7B、InternVL3-8B 和 MiniCPM-V-4.6。结果显示，二分类危险存在准确率（88.0%-89.2%）与多数类基线（89.0%）基本持平，而细粒度推理明显薄弱：危险类别准确率低于 40%，风险时机准确率仅 1.6%-12.4%，防御动作精确匹配低于 2%。
正文节选
BD-HazardVLM: Probing Vision-Language Models for Latent Defensive-Driving Hazards in Bangladesh Road Scenes Abstract Vision-language models are increasingly used for driving-scene understanding, yet their defensive reasoning under weakly structured Global South traffic remains unclear. We introduce BD-HazardVLM, a diagnostic benchmark of 500 Bangladesh road-scene images—375 from RSUD20K and 125 from TFP-BD—annotated for hazard presence, primary hazard category, risk timing, severity, defensive a