TerraLogic: A Benchmark for Hierarchical Geospatial Reasoning in Earth Observation

2026-07-14

AI Summary

This study addresses the limited scope of existing remote sensing research, which predominantly focuses on perceptual tasks while lacking systematic evaluation of hierarchical geospatial reasoning at the cognitive level. To bridge this gap, we propose TerraLogic—the first hierarchical benchmark specifically designed for cognitive-level geospatial reasoning—comprising 545 scene-driven Earth observation tasks. We further introduce HieraPlan, an agent architecture that integrates functionally layered tool organization, fault-tolerant reasoning mechanisms, and multimodal remote sensing data (optical, SAR, and infrared) to enable structured abstraction and long-horizon planning. Experimental results demonstrate that current methods exhibit limited performance on this benchmark, whereas HieraPlan substantially enhances reasoning capability, cross-modal generalization, and robustness in error recovery.

Abstract

Beyond perception, reasoning is essential in remote sensing for advanced interpretation, inference, and decision-making. Recent advances in large language models (LLMs) have enabled tool-augmented agents that leverage external tools to perform complex analytical tasks. However, existing studies in remote sensing primarily focus on perception-oriented tasks, leaving cognitive geospatial reasoning largely underexplored. To address this gap, we introduce TerraLogic, a benchmark for geospatial reasoning. TerraLogic comprises 545 scenario-driven, hierarchy-aware tasks, such as hazard vulnerability assessment, urban heat island analysis, and forest fragmentation dynamics, spanning optical, Synthetic Aperture Radar (SAR), and infrared (IR) imagery. It advances evaluation beyond recognition and monitoring toward cognitive-level geospatial analysis. To facilitate evaluation on TerraLogic, we further propose HieraPlan, a tool-augmented agent that organizes toolkits into functional hierarchies and performs fault-tolerant reasoning. HieraPlan enables structured abstraction, robust recovery from tool failures, and stable long-horizon planning. Extensive experiments demonstrate that current approaches struggle with hierarchical geospatial reasoning, while HieraPlan provides a strong baseline with improved reasoning, cross-modal generalization, and error handling. The dataset and agent code are publicly available at https://github.com/Ireliya/TerraLogic.

Problem

• geospatial reasoning
• hierarchical reasoning
• remote sensing
• Earth observation
• cognitive reasoning

Innovation

• hierarchical geospatial reasoning
• tool-augmented agent
• Earth observation benchmark
• cross-modal generalization
• fault-tolerant planning