EO-Agents: A Three-Agent LLM Pipeline for Earth Observation Hypothesis Generation

2026-07-01

AI Summary

Traditional scientific hypothesis generation relies on unstructured text, making it difficult to directly link with Earth observation data. This work proposes the first framework that integrates knowledge graphs with a multi-agent large language model, leveraging NASA’s Earth Observing Knowledge Graph. By employing a heterogeneous graph neural network, the framework recommends high-quality combinations of datasets and utilizes a three-agent pipeline to automatically generate, filter, and evaluate structured scientific hypotheses. Applied to 1,475 NASA datasets, the method produced 160 cross-domain hypotheses whose novelty and validity are comparable to those co-used in real research, demonstrating its capacity for data-driven, interpretable, and effective hypothesis discovery.

Abstract

Large language models have recently been explored for scientific hypothesis generation, but most prior work relies on unstructured literature and free-form textual claims. We present a pipeline for Earth observation that grounds hypothesis generation directly in the NASA Earth Observation Knowledge Graph. A heterogeneous graph neural network trained on historical co-usage relations ranks candidate dataset pairings, and a three-agent LLM pipeline filters, generates, and evaluates structured research hypotheses. Applied to 1,475 NASA datasets, the system produces 160 hypotheses spanning multiple Earth-science domains, including ecohydrology, glaciology, aerosol--cloud interactions, vegetation phenology, and stratospheric chemistry. Model-predicted novel dataset pairings are rated nearly as plausible as held-out real co-usages from the literature, indicating that the pipeline surfaces scientifically coherent yet unexplored combinations. A 2*2*2 factorial experiment across GPT-5.2 and Claude Sonnet 4.6 shows that hypothesis rankings remain stable, while absolute scores depend strongly on judge identity, highlighting limitations of single-judge LLM evaluation.

Problem

• hypothesis generation
• Earth observation
• knowledge graph
• dataset pairing
• scientific discovery

Innovation

• three-agent LLM pipeline
• Earth Observation Knowledge Graph
• heterogeneous graph neural network
• structured hypothesis generation
• dataset co-usage ranking