MSEarth: A Multimodal Benchmark for Earth Science Phenomenon Discovery with MLLMs

TL;DR

MSEarth is a comprehensive multimodal benchmark dataset designed to evaluate large language models' scientific reasoning in earth science, covering five spheres with over 289,000 figures and supporting various reasoning tasks.

Contribution

The paper introduces MSEarth, a high-quality, open-access multimodal dataset and benchmark for earth science, enabling better evaluation of MLLMs' scientific reasoning capabilities.

Findings

MSEarth includes over 289,000 figures with detailed captions and reasoning.

The benchmark supports figure captioning, multiple choice, and open-ended reasoning tasks.

It provides a scalable resource for developing earth science MLLMs.

Abstract

The rapid advancement of multimodal large language models (MLLMs) offers new opportunities for complex scientific challenges, yet their application in earth science-especially at the graduate level-remains underexplored due to a lack of benchmarks reflecting the depth and complexity of geoscientific reasoning. Existing datasets often rely on synthetic data or simple figure-caption pairs, failing to capture the nuanced reasoning required for real-world applications. To address this, we introduce MSEarth, a multimodal scientific dataset and benchmark curated from high-quality, open-access publications. Covering the five major spheres of Earth science-atmosphere, cryosphere, hydrosphere, lithosphere, and biosphere-MSEarth features over 289K figures with refined captions enriched by contextual discussions and reasoning from the original papers. The benchmark supports tasks such as scientific figure captioning, multiple choice questions, and open-ended reasoning, providing a scalable, high-fidelity resource for developing and evaluating MLLMs in scientific reasoning.