ResearchAI Research

EO-Agents Generate Earth Observation Hypotheses with LLMs

Mahyar Ghazanfari, Amin Tabrizian, Armin Mehrabian, Peng Wei· July 3, 2026 View original

Key takeaways

  • EO-Agents use a three-agent LLM pipeline for Earth observation hypothesis generation.
  • It grounds hypotheses in the NASA Earth Observation Knowledge Graph.
  • The system identifies novel, plausible dataset pairings for research.
  • It significantly aids scientific discovery across various Earth-science domains.

Who benefits

Environmental ScienceClimate ResearchSpace ExplorationAgricultureDisaster Management

Summary

A three-agent LLM pipeline, EO-Agents, generates structured Earth observation research hypotheses by grounding them in the NASA Earth Observation Knowledge Graph. The system ranks dataset pairings and uses LLMs to filter, generate, and evaluate hypotheses across various Earth-science domains.

While Large Language Models (LLMs) have been explored for scientific hypothesis generation, much of the prior work relies on unstructured text. A new pipeline, EO-Agents, grounds hypothesis generation directly in the NASA Earth Observation Knowledge Graph, offering a more structured approach for Earth science. The system first uses a heterogeneous graph neural network, trained on historical co-usage relations, to rank candidate dataset pairings. Following this, a three-agent LLM pipeline takes over to filter, generate, and evaluate structured research hypotheses. Applied to 1,475 NASA datasets, EO-Agents produced 160 hypotheses spanning diverse Earth-science domains, including ecohydrology and glaciology. The model-predicted novel dataset pairings were rated nearly as plausible as actual co-usages from literature, indicating its ability to surface scientifically coherent yet unexplored combinations.

Why it matters

This framework significantly accelerates scientific discovery in Earth observation by automating the generation of plausible, novel research hypotheses, potentially leading to breakthroughs in understanding climate change and environmental phenomena.

How to implement this in your domain

  1. 1Explore integrating knowledge graph-grounded LLM pipelines for hypothesis generation in your research domain.
  2. 2Identify and structure relevant domain-specific knowledge graphs for LLM interaction.
  3. 3Develop multi-agent LLM architectures for filtering, generating, and evaluating complex scientific claims.
  4. 4Pilot the system on internal datasets to identify novel correlations or research avenues.
  5. 5Collaborate with domain experts to validate and refine LLM-generated hypotheses.

Original post by Mahyar Ghazanfari, Amin Tabrizian, Armin Mehrabian, Peng Wei

"arXiv:2607.01584v1 Announce Type: new 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 hyp…"

View on X

Originally posted by Mahyar Ghazanfari, Amin Tabrizian, Armin Mehrabian, Peng Wei on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses