Since large language models (LLMs) generally have a large parameter scale and acquire knowledge across multidisciplinary corpora, their parameters tend to exhibit varying levels of importance for GIS knowledge alignment. Specifically, certain model layers of LLMs may already align well with GIS knowledge, while others may require significant adaptation and thus play a more critical role during the fine-tuning process. However, representative fine-tuning methods, such as low-rank adaptation (LoRA), apply a uniform treatment to all parameters, overlooking their varying importance for GIS downstream tasks. This strategy may disrupt well-adapted layers or limit less-adapted ones from effectively encoding GIS knowledge. Therefore, we develop a novel GIS knowledge-aware LoRA allocation (GLoRA) scheme that adaptively allocates trainable parameters to layers based on their importance. Specifically, GLoRA first investigates a GIS knowledge-aware strategy to identify the importance of model layers for a given GIS task in a data-dependent way. Second, it dynamically adjusts the size of LoRA modules, allocating more parameters to more important layers to enhance their representation capability. This adaptive approach ensures efficient parameter utilization while preserving the strengths of well-adapted layers. We evaluated GLoRA on three GIS-related tasks, and the results show its improved performance compared to recent baselines with comparable parameter budgets.

Large Language Models (LLMs) excel in natural language-relevant tasks like text generation and question answering Q&A. To further expand their application, efforts focus on enabling LLMs to utilize real-world tools. However, their tool-use ability in professional GIS remains under explored due to two main challenges. Firstly, LLMs are usually trained on general-domain corpora, lacking sufficient and comprehensive GIS-specific data to align with professional knowledge, including understanding the functions of GIS tools. Secondly, researchers often need to combine multiple GIS tools to solve geospatial tasks. To address these challenges, we propose a trainable method to enable LLMs to master GIS tools. We curated a comprehensive set of resources: instruction-response data (GeoTool, 1950 instructions) to enhance the understanding of LLMs for GIS tools, instruction-solution data (GeoSolution, 3645 instructions) to improve their ability to generate tool-use solutions for geospatial tasks, and annotated instruction-solution evaluation data (GeoTask, 300 instructions) for evaluating LLMs’ GIS tool-use proficiency. Using the collected training data (GeoTool and GeoSolution), we fine-tuned a professional-domain LLM called GeoTool-GPT based on an open-source general-domain LLM, the LLaMA-2-7b model. The experiment based on evaluation data validates our method’s effectiveness in enhancing the tool-use ability of general-domain LLMs in the professional GIS domain, with the performance of our model closely approaching that of GPT-4.

Intelligent map analysis is an important yet challenging topic. Recently, the development of large models, especially Visual Language Models (VLMs), has shown potential for intelligent image analysis. However, these models are primarily trained on natural images, which have intrinsic differences from maps. Consequently, there remains a gap in applying existing general-domain VLMs to map analysis. To address this issue, we propose a framework for developing a specialized VLM, called MapReader. To achieve this goal, a comprehensive data resource is collected using a strategy that combines self-instruct with expert refinement, including training data (MapTrain: 2,000 pairs of maps and descriptions) and evaluation data (MapEval: 250 maps and 500 map-related questions). Based on the training data, MapReader is fine-tuned on top of a general-domain VLM to learn to understand and describe map contents. The evaluation results on MapEval suggest that: (1) MapReader can accept map inputs and generate detailed descriptions of core geographic information, and it also possesses visual question-answering capabilities, showing potential for application in various map analysis scenarios, such as accessible map reading and robotic map usage; (2) The proposed data collection strategy is effective, and the collected dataset can serve as a benchmark to promote further map analysis research.

Bike-sharing systems become more and more popular in the urban transportation system, because of their convenience in recent years. However, due to the high daily usage and lack of effective maintenance, the number of bikes in good condition decreases significantly, and vast piles of broken bikes appear in many big cities. As a result, it is more difficult for regular users to get a working bike, which causes problems both economically and environmentally. Therefore, building an effective broken bike prediction and recycling model becomes a crucial task to promote cycling behavior. In this paper, we propose a predictive model to detect the broken bikes and recommend an optimal recycling program based on the large scale real-world sharing bike data. We incorporate the realistic constraints to formulate our problem and introduce a flexible objective function to tune the trade-off between the broken probability and recycled numbers of the bikes. Finally, we provide extensive experimental results and case studies to demonstrate the effectiveness of our approach.