arXiv 2605.122372026-09-30cs.CV

UHR-Micro: 诊断和缓解地球观测VLMs中的分辨率错觉

UHR-Micro: Diagnosing and Mitigating the Resolution Illusion in Earth Observation VLMs

National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing（国家空间智能信息处理科技重点实验室）；Beijing Institute of Technology（北京理工大学）；School of Computer Science（计算机学院）；Wuhan University（武汉大学）；Zhongguancun Academy（中关村学院）；State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing（测绘遥感信息工程国家重点实验室）；Hong Kong Polytechnic University（香港理工大学）

Shuo Ni, Tong Wang, Jing Zhang, Di Wang, He Chen, Haonan Guo, Ning Zhang, Bo Du

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

UHR-Micro通过11253条指令和1212张超高清图像，评估VLM在高分辨率地球观测图像中的空间极限表现，揭示高分辨率输入下空间定位和证据解析的失败，并提出MAP代理以证据为中心提升微尺度感知。

摘要：

视觉-语言模型（VLMs）越来越多地处理超高清（UHR）地球观测图像，但它们仍易受大尺度场景上下文与微尺度目标之间的严重尺度不匹配影响。我们将其称为“分辨率错觉”：更高的输入分辨率提供了更丰富的视觉细节，但并不一定产生可靠的对空间小、任务相关证据的感知。为衡量这一挑战，我们引入UHR-Micro，一个包含11,253条指令和1,212张UHR图像的基准，旨在评估VLM在原生地球观测图像空间极限下的表现。UHR-Micro涵盖多样化的微目标尺度、上下文需求、任务家族和视觉条件，并提供诊断注释以支持受控评估和细粒度错误归因。对代表性高分辨率VLM的实验显示，尽管有高分辨率输入，仍存在显著的空间定位和证据解析失败。进一步分析表明，这些失败并非完全由增加模型容量解决，而是与定位和使用任务相关微证据的指导不足密切相关。受此发现启发，我们提出Micro-evidence Active Perception（MAP），一个参考代理，将查询分解为证据寻找步骤，主动检查候选区域，并将其回答基于局部观察。MAP代理通过使高分辨率推理以证据为中心而非图像为中心，提高了微尺度感知。UHR-Micro和MAP代理共同提供了一个诊断平台，用于评估、理解和推进地球观测VLM中的高分辨率推理。数据集和源代码发布于https://github.com/MiliLab/UHR-Micro。

英文摘要：

Vision-Language Models (VLMs) are increasingly used to analyze ultra-high-resolution (UHR) Earth observation imagery, yet they face a severe scale mismatch between broad scene context and micro-scale targets. We refer to this phenomenon as a "resolution illusion": higher input resolution provides access to more visual detail, but does not necessarily translate into reliable perception of task-relevant micro-evidence. To benchmark this challenge, we introduce UHR-Micro, a benchmark comprising 11,072 instructions grounded in 1,212 UHR images, designed to evaluate VLMs on micro-scale evidence in native-resolution Earth observation imagery. UHR-Micro spans diverse target scales, task families, and visual conditions, with each sample having an objectively verifiable target. Experiments with representative high-resolution VLMs show substantial failures in localizing and interpreting task-relevant evidence, despite access to high-resolution inputs. Further analysis shows that model scaling alone is insufficient, while localized evidence substantially improves performance, pointing to evidence access as a major bottleneck. Motivated by this finding, we propose Micro-evidence Active Perception (MAP), which constructs a spatially grounded evidence state from localized observations and supplements it when the current evidence is insufficient. Across two backbone VLMs, MAP improves UHR-Micro performance by 8.65 percentage points on average. UHR-Micro and MAP provide a framework for diagnosing and improving high-resolution reasoning in Earth observation VLMs. Datasets and source code were released at https://github.com/MiliLab/UHR-Micro.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。