Scopus

Fine-Tuning Generative LLMs for Geographic Analysis: Comparing GPT-4o, Gemma-2B, and DeepSeek-R1

Năm XB 2026 Tạp chí / Hội thảo International Conference on Advances in Information and Communication Technology (ICTA2025) Vol 2, 308–316 Đơn vị CNTT DOI / Link https://doi.org/10.1007/978-3-032-18159-6_33 ↗

Tác giả

Kim-Son Nguyen ; The-Vinh Nguyen ; Huu-Huy Ngo ; Van-Viet Nguyen ; Thi Minh-Hue Luong ; Huu-Khanh Nguyen ; Duc-Binh Nguyen ✉

Tóm tắt

The integration of Large Language Models (LLMs) into geospatial analysis presents significant opportunities for advancing spatial intelligence. However, questions remain regarding the optimal balance between predictive performance and computational efficiency. This study provides a comprehensive investigation into modern LLMs for geospatial reasoning tasks. We compare the state-of-the-art proprietary model GPT-4o against two fine-tuned lightweight open-weight models: Gemma-2B and DeepSeek-R1. To ensure robust evaluation, we conduct experiments on the complete 100,000-prompt GeoLLM benchmark, using the original fine-tuned LLaMA-7B as our primary baseline. Our experimental results demonstrate that GPT-4o achieves a new performance record with an average Pearson's r2 of 0.862. More significantly, the fine-tuned Gemma-2B model surpasses the larger LLaMA-7B baseline while …

Tài liệu tham khảo

[1] Li, Z., Ning, H.: Autonomous GIS: the next-generation AI-powered GIS. Int. J. Digital Earth 16(2), 4668–4686 (2023)

[2] Diehr, J., Ogunyiola, A., Dada, O.: Artificial intelligence and machine learning-powered GIS for proactive disaster resilience in a changing climate. Ann. GIS 31(2), 287–300 (2025)

[3] Mai, G., Hu, H., Liu, Y., Janowicz, K.: On the opportunities and challenges of foundation models for GeoAI. ACM Trans. Spatial Algo. Syst. 10(2), 1–46 (2024)

[4] AlZu’bi, S., Alsmirat, Y., Al-Betar, M.: Exploring the capabilities and limitations of ChatGPT and alternative big language models. Artif. Intell. Appl. 2(1), 28–37 (2024)

[5] Zhang, Y., Jiang, M., Zhang, Y.: GeoGPT: an assistant for understanding and processing geospatial tasks. Int. J. Appl. Earth Observ. Geoinform. 131, Art. no. 103976 (2024)

[6] Zhang, Y., et al.: BB-GeoGPT: a framework for learning a large language model for geographic information science. Preprint submitted to Elsevier (2023)

[7] Manvi, R., Tan, Y., Tang, B., Zhang, A.: GeoLLM: extracting geospatial knowledge from large language models. In: Proceedings of the International Conference on Learning Representations (ICLR) (2024)

[8] Feng, Y., Zhou, Y., Zhang, H.: GeoQAMap – geographic question answering with maps leveraging LLM and open knowledge base. In: Proceedings of GIScience 2023 (2023)

[9] Ning, H., et al.: An autonomous GIS agent framework for geospatial data retrieval. Int. J. Dig. Earth 18(1), Art. no. 2458688 (2024)

[10] Fulman, N., Deb, A., Curtis, K.: Distortions in judged spatial relations in large language models. Prof. Geogr. 76(6), 703–711 (2024)

[11] Hatvani, M., et al.: Essential questions in earth and geosciences according to large language models. Geosci. Front. 16, Art. no. 101696 (2024)

[12] Lin, J., Zhang, T., Gupta, A.: A Systematic Review of Geospatial Location Embedding Approaches in Large Language Models. arXiv preprint arXiv:2306.05429 (2023)

[13] Memduhoğlu, A., Doganay, A., Turker, M.: Enriching building function classification using large language model embeddings of OpenStreetMap tags. In: Proceedings of the 31st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, pp. 533–536 (2023)

[14] Li, W., Batty, M., Goodchild, M.F.: Real-Time GIS for smart cities. Int. J. Geogr. Inf. Sci. 34(3), 311–324 (2019)

[15] Akinboyewa, T., Li, Z., Ning, H., Lessani, M.N.: GIS copilot: towards an autonomous GIS agent for spatial analysis. Int. J. Dig. Earth 18(1), 2497489 (2025)

← Quay lại danh sách bài báo