The Geologic Map Intelligent Platform: AI-Enabled Digital Transformation and Building a Multimodal Application Ecosystem

Geologic maps are undergoing a paradigm shift from static illustrations to dynamic, intelligent knowledge platforms. Traditionally, geologic maps have served specialized fields in fixed image formats. However, their closed information systems, weak interactivity, and difficulties in cross-domain integration have limited the full release of their value. In recent years, advancements in Artificial Intelligence (AI) and Multimodal Large Language Models (MLLMs) have provided a new pathway for the digital reconstruction and intelligent application of geologic maps.

In collaboration with Microsoft Research Asia, our project team has proposed and constructed an open, extensible intelligent platform for geologic map comprehension and service. This platform is based on high-quality digitized geologic map datasets and utilizes MLLMs to achieve semantic parsing, knowledge association, and natural language interaction with geologic maps. The established platform not only supports the accurate identification and extraction of fundamental map elements (such as legends, lithology, and structures) but also enables the following multi-level application scenarios:

• Intelligent Interaction and Q&A: Users can directly query geologic information using natural language—for example, "What faults are distributed in this area?" or "What is the formation age of a certain rock layer?" The system generates accurate answers by integrating graphic-text information and domain knowledge.
• Scientific Research and Educational Tools: It provides an interactive, annotatable interface for geologic map learning, supporting classroom teaching, professional training, and interdisciplinary research.

The platform is supported by core technologies including the first-ever multimodal benchmark for geologic map understanding, GeoMap-Bench, and the intelligent agent framework, GeoMap-Agent, which significantly outperforms general-purpose vision-language models on multiple tasks. Geologic maps are no longer merely "base maps" or "reference maps"; they have become an intelligent knowledge base connecting geologic data, professional expertise, and multi-domain applications.

Looking ahead, the geologic map platform will further integrate real-time sensor data, remote sensing information, and socio-economic factors, driving the earth sciences towards a new era characterized by openness, collaboration, and intelligence. It is poised to play a central role in scientific discovery, engineering safety, sustainable resource utilization, and the building of societal resilience.