Researchers have developed an agentic search system designed to improve the discovery of Earth observation data from NASA's extensive archives. This system leverages large language models and knowledge graphs to interpret natural language queries and identify relevant datasets and tools. To evaluate its effectiveness, an open benchmark called NASA-EO-Bench was created, comprising 47,000 query-dataset pairs. The system demonstrated significant improvements in retrieval accuracy, outperforming traditional methods and further enhanced by an agentic reranking stage. AI
IMPACT This agentic search system could streamline scientific research by making vast datasets more accessible, potentially accelerating discoveries in Earth sciences.
RANK_REASON The cluster contains a research paper detailing a new system and benchmark. [lever_c_demoted from research: ic=1 ai=1.0]
- Giovanni
- Harmony
- Hugging Face
- NASA
- NASA Earth Observation Knowledge Graph
- NASA-EO-Bench
- Science Discovery Engine
- Worldview
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →