EO-Gym: A Multimodal, Interactive Environment for Earth Observation Agents

TL;DR

EO-Gym introduces a comprehensive, interactive environment for Earth Observation analysis, enabling multimodal, tool-using agents to perform complex reasoning across spatial, temporal, and sensor modalities.

Contribution

It provides a novel controlled framework and benchmark for interactive EO reasoning, addressing limitations of existing fixed-input, single-turn tasks.

Findings

Strong models struggle with interactive EO reasoning, especially across time and modalities.

Fine-tuned EO-Gym-4B improves Pass@3 from 0.49 to 0.74.

EO-Gym offers a reproducible environment for developing and evaluating EO agents.

Abstract

Earth Observation (EO) analysis is inherently interactive: resolving uncertainty often requires expanding the region of interest, retrieving historical observations, and switching across sensors such as optical and Synthetic Aperture Radar. However, most EO benchmarks collapse this process into fixed-input, single-turn tasks. To address this gap, we present EO-Gym, a controlled executable framework for multimodal, tool-using EO agents that formulates EO analysis as a Gymnasium-style local geospatial workspace backed by more than 660k multimodal files indexed by location, time, and sensor type, with 35 EO-specialized tools spanning six task families. Built on this environment, we construct EO-Gym-Data, a benchmark of 9,078 trajectories and 34,604 reasoning steps, and grounded in eight public EO datasets together with Landsat and Sentinel-2 imagery. Evaluating $10$ open and closed VLMs shows that strong general-purpose models still struggle with interactive EO reasoning, especially on temporal and cross-modal workflows. As a reference baseline, EO-Gym-4B, obtained by fine-tuning Qwen3-VL-4B-Instruct on EO-Gym-Data, improves overall Pass@3 from $0.49$ to $0.74$ under the main evaluation setting. O-Gym provides a reproducible environment for interactive EO agents, operationalizing EO as an evidence-gathering problem that requires planning across geospatial, temporal, and sensing modality.