AEFEcon-GPT: Fact-Grounded Language Generation from Multiscale Geospatial Foundation Representations
Authors:
- Shuyang HouSHShuyang Hou
- Guanyu ChenGCGuanyu Chen
- Haoyue JiaoHJHaoyue Jiao
- Lutong XieLXLutong Xie
- Ziqi LiuZLZiqi Liu
- Donghai LiuDLDonghai Liu
- Shaowen WuSWShaowen Wu
- Yaxian QingYQYaxian Qing
- Huayi WuHuayi Wu
Preprint
- 27 October 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Geospatial foundation models encode geographic environments into compact representations, but their interpretation across spatial scales and time remains challenging. We investigate how frozen AlphaEarth Foundations (AEF) representations can generate descriptions of land-surface characteristics, urban functions, and socioeconomic conditions. We develop AEFEcon-GPT, which aligns multiscale AEF representations with a language model through a learned projection and two-stage semantic alignment. Structured labels from independent geographic datasets supervise regional types, attributes, and spatial–temporal relations; auxiliary attributes are not required during inference. The framework supports static description and spatial comparison at pixel, area-of-interest, and 1 km grid scales, and temporal interpretation at the grid scale. Experiments cover 36 Chinese cities during 2017–2024, with 167,088 supervision instances and 3,282 evaluation samples. Static interpretation achieves 0.620 exact-grade accuracy, 0.858 adjacent-grade accuracy, and 0.934 attribute coverage. Attribute-level accuracies are 0.608 for temporal change directions and 0.625 for spatial comparison relations. Representation diagnostics and ablations show that factual descriptions depend on both information encoded in AEF representations and structured semantic supervision. The framework enables verifiable geographic interpretation of geospatial foundation representations across scales and time.Keywords
This publication has 15 references indexed in Scilit:
- S2Vec: Self-Supervised Geospatial Embeddings for the Built EnvironmentACM Transactions on Spatial Algorithms and Systems, 2026
- RSGPT: A remote sensing vision language model and benchmarkISPRS Journal of Photogrammetry and Remote Sensing, 2025
- UrbanGPT: Spatio-Temporal Large Language Models2024
- UrbanCLIP: Learning Text-enhanced Urban Region Profiling with Contrastive Language-Image Pretraining from the Web2024
- SpectralGPT: Spectral Remote Sensing Foundation ModelIEEE Transactions on Pattern Analysis and Machine Intelligence, 2024
- SkyScript: A Large and Semantically Diverse Vision-Language Dataset for Remote SensingProceedings of the AAAI Conference on Artificial Intelligence, 2024
- GeoVeX2023
- Beyond the First Law of Geography: Learning Representations of Satellite Imagery by Leveraging Point-of-Interests2022
- hex2vec2021
- Urban2Vec: Incorporating Street View Imagery and POIs for Multi-Modal Urban Neighborhood EmbeddingProceedings of the AAAI Conference on Artificial Intelligence, 2020