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

TL;DR

This paper introduces UHR-Micro, a benchmark and diagnostic platform for evaluating and improving high-resolution Earth observation vision-language models, addressing the challenge of the resolution illusion where higher resolution does not guarantee better micro-scale perception.

Contribution

The paper presents UHR-Micro, a comprehensive benchmark for micro-level reasoning in Earth observation VLMs, and proposes MAP, an active perception agent that enhances micro-evidence grounding.

Findings

High-resolution VLMs often fail in spatial grounding despite detailed inputs.

Increasing model capacity does not fully resolve micro-evidence perception issues.

MAP improves micro-level perception by actively seeking and grounding evidence.

Abstract

Vision-Language Models (VLMs) increasingly operate on ultra-high-resolution (UHR) Earth observation imagery, yet they remain vulnerable to a severe scale mismatch between large-scale scene context and micro-scale targets. We refer to this empirical gap as a "resolution illusion": higher input resolution provides the appearance of richer visual detail, but does not necessarily yield reliable perception of spatially small, task-relevant evidence. To benchmark this challenge, we introduce UHR-Micro, a benchmark comprising 11,253 instructions grounded in 1,212 UHR images, designed to evaluate VLMs at the spatial limits of native Earth observation imagery. UHR-Micro spans diverse micro-target scales, context requirements, task families, and visual conditions, and provides diagnostic annotations that support controlled evaluation and fine-grained error attribution. Experiments with representative high-resolution VLMs show substantial failures in spatial grounding and evidence parsing, despite access to high-resolution inputs. Further analysis suggests that these failures are not fully resolved by increasing model capacity, but are closely tied to insufficient guidance in locating and using task-relevant micro-evidence. Motivated by this finding, we propose Micro-evidence Active Perception (MAP), a reference agent that decomposes queries into evidence-seeking steps, actively inspects candidate regions, and grounds its answers in localized observations. MAP-Agent improves micro-level perception by making high-resolution reasoning evidence-centered rather than image-centered. Together, UHR-Micro and MAP-Agent provide a diagnostic platform for evaluating, understanding, and advancing high-resolution reasoning in Earth observation VLMs. Datasets and source code were released at https://github.com/MiliLab/UHR-Micro.