LLM4ST: Leveraging large language models for multimodal spatiotemporal modeling

2026-08-01

PRE

Zhenzhen Zhao

Guojiang Shen

Yinghui Liu

Qihong Pan

Xiangfan Chen

Renhe Jiang

Xiangjie Kong *

DOI：10.1016/j.future.2026.108736

Original

Original request for help

Save

Abstract

En 中文

• LLM4ST introduces an LLM-based multimodal framework for spatiotemporal modeling. • The model fuses tokens, language descriptions, and backbone embeddings effectively. • A cross-modal transformer enables joint reasoning over numeric and semantic cues. • The experiment results demonstrate the effectiveness of the framework.

LLM4ST: Leveraging large language models for multimodal spatiotemporal modeling

Abstract

Journal

Organization

Cited Papers

Citing Papers