Drone-PUR：无人机场景视觉线索发现与推理基准
原标题：Drone-PUR: From Visual Clue Discovery to Understanding and Reasoning in UAV Scenes
AI 摘要
天津大学等机构的研究者提出 Drone-PUR，一个面向无人机场景的视觉线索发现、理解与推理数据集和基准，包含 11 项任务、17.5 万训练样本和 2 万测试样本，覆盖单图、图像序列、跨模态图像和多视角。评测 22 个通用视觉语言模型后发现，它们在空间判断、跨模态与跨视角推理上存在持续局限，且感知能力更强的模型推理表现更好。团队还基于 Qwen3-VL 分阶段微调得到 UAV-PUR，其 4B 和 8B 版本的多选题准确率分别比预训练版本提升 18.33 和 17.52 个百分点。
正文节选
Drone-PUR: From Visual Clue Discovery to Understanding and Reasoning in UAV ScenesThanks: This work was supported in part by Science and Technology Innovation Program of Xiongan New Area (Grant No.2025XAGG0039), in part by the National Natural Science Foundation of China (NSFC) under Grant 62436002, and in part by Tianjin Natural Science Funds for Distinguished Young Scholar under Grant 23JCJQJC00270. (Corresponding author: Pengfei Zhu.) Jiahe Wu, Bing Cao, Ruipu Zhao, Jiekang Feng and Qinghua H