GeoEyes: On-Demand Visual Focusing for Evidence-Grounded Understanding of Ultra-High-Resolution Remote Sensing Imagery

TL;DR

GeoEyes introduces a staged training framework with reinforcement learning to improve evidence gathering in ultra-high-resolution remote sensing visual question answering, significantly enhancing accuracy by enabling effective zooming strategies.

Contribution

It presents a novel staged training approach and reinforcement learning method to address tool usage homogenization in UHR remote sensing VQA.

Findings

Achieves 54.23% accuracy on XLRS-Bench.

Effectively learns on-demand zooming with proper stopping behavior.

Substantially improves evidence acquisition in UHR remote sensing tasks.

Abstract

The "thinking-with-images" paradigm enables multimodal large language models (MLLMs) to actively explore visual scenes via zoom-in tools. This is essential for ultra-high-resolution (UHR) remote sensing VQA, where task-relevant cues are sparse and tiny. However, we observe a consistent failure mode in existing zoom-enabled MLLMs: Tool Usage Homogenization, where tool calls collapse into task-agnostic patterns, limiting effective evidence acquisition. To address this, we propose GeoEyes, a staged training framework consisting of (1) a cold-start SFT dataset, UHR Chain-of-Zoom (UHR-CoZ), which covers diverse zooming regimes, and (2) an agentic reinforcement learning method, AdaZoom-GRPO, that explicitly rewards evidence gain and answer improvement during zoom interactions. The resulting model learns on-demand zooming with proper stopping behavior and achieves substantial improvements on UHR remote sensing benchmarks, with 54.23% accuracy on XLRS-Bench.