GeoGPT.RAG Technical Report

2025-08-18

AI Summary

Large language models (LLMs) suffer from insufficient domain expertise in Earth sciences and struggle to integrate user-specific literature. To address this, we propose GeoGPT—an open-source retrieval-augmented generation (RAG) system tailored for geoscience research. Methodologically, we construct a specialized geoscience knowledge base and introduce a domain-adaptive retrieval framework comprising two novel components: GeoEmbedding, a lightweight, open-source geoscience-specific embedding model; and GeoReranker, a fine-grained re-ranking model—both domain-finetuned to enhance semantic alignment and retrieval accuracy. The system further supports dynamic construction of personalized knowledge bases via user-uploaded documents. Experiments demonstrate significant improvements in both answer accuracy and factual consistency for geoscience question answering. Core modules are publicly released under open-source licenses, fostering transparency, reproducibility, and collaborative advancement of AI tools in Earth sciences.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

GeoGPT is an open large language model system built to advance research in the geosciences. To enhance its domain-specific capabilities, we integrated Retrieval Augmented Generation(RAG), which augments model outputs with relevant information retrieved from an external knowledge source. GeoGPT uses RAG to draw from the GeoGPT Library, a specialized corpus curated for geoscientific content, enabling it to generate accurate, context-specific answers. Users can also create personalized knowledge bases by uploading their own publication lists, allowing GeoGPT to retrieve and respond using user-provided materials. To further improve retrieval quality and domain alignment, we fine-tuned both the embedding model and a ranking model that scores retrieved passages by relevance to the query. These enhancements optimize RAG for geoscience applications and significantly improve the system's ability to deliver precise and trustworthy outputs. GeoGPT reflects a strong commitment to open science through its emphasis on collaboration, transparency, and community driven development. As part of this commitment, we have open-sourced two core RAG components-GeoEmbedding and GeoReranker-to support geoscientists, researchers, and professionals worldwide with powerful, accessible AI tools.

Problem

• Advancing geoscience research with open large language model
• Enhancing domain-specific capabilities using Retrieval Augmented Generation
• Improving retrieval quality through fine-tuned embedding and ranking models

Innovation

• Integrated Retrieval Augmented Generation for geosciences
• Fine-tuned embedding and ranking models for relevance
• Open-sourced GeoEmbedding and GeoReranker components