In the field of geoscience, geological maps are crucial tools for understanding the Earth's surface and subsurface structures. However, interpreting these complex diagrams requires specialized knowledge and extensive experience. To enhance intelligence in this field, Microsoft Research Asia recently introduced GeoMap-Bench, a new benchmark designed to evaluate the performance of multi-modal large language models (MLLMs) in geological map understanding.
The launch of GeoMap-Bench marks a significant step forward in AI applications for geological map interpretation. Microsoft researchers collaborated with experts from the Chinese Academy of Geological Sciences and Wuhan University to identify five key capabilities required for geological map understanding: information extraction, localization, reference resolution, reasoning, and analysis. These capabilities span from acquiring basic information to performing complex logical tasks, ensuring AI can comprehensively understand geological maps.
Researchers used over 7,000 geological maps to create GeoMap-Bench, generating more than 3,000 questions with standard answers to assess the limitations of existing multi-modal large language models in geological map interpretation. Common challenges include processing high-resolution images, the need for specialized domain knowledge, and diverse visual representations.
To further advance the intelligent interpretation of geological maps, Microsoft also developed GeoMap-Agent. This intelligent agent not only possesses the capability to digitize geological maps but also shows significant improvements in high-resolution image processing and domain knowledge application. The architecture of GeoMap-Agent comprises three core modules: hierarchical information extraction, domain knowledge injection, and enhanced prompt-based question answering. The combination of these modules enables AI to efficiently and accurately analyze complex geological information.
Researchers state that the introduction of GeoMap-Bench and GeoMap-Agent provides a new research foundation for AI applications in geology and will significantly promote the digitalization and intelligence of geological information, bringing profound impacts to broader fields such as disaster warning and resource exploration.
