Related Experiment Video
Updated: May 28, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
A scalable and generic framework for city-wide traffic prediction with large language model
Jinlei Zhang1,2, Congwang Deng3, Lixing Yang4,5
1School of Systems Science, Beijing Jiaotong University, Beijing, China. zhangjinlei@bjtu.edu.cn.
Nature Communications
|May 26, 2026
Summary
A new large language model framework (LLM-UTP) offers scalable, city-wide urban traffic prediction across diverse modes and scenarios. This advancement improves traffic management for future smart cities.
Area of Science:
- Intelligent Transportation Systems
- Artificial Intelligence
- Urban Planning
Background:
- Future urban transportation demands integrated, scalable traffic prediction frameworks.
- Existing models lack generalizability across transport modes and scenarios.
Purpose of the Study:
- To propose a large language model (LLM)-based framework (LLM-UTP) for scalable and generic city-wide urban traffic prediction.
- To address limitations in current traffic flow prediction models.
Main Methods:
- Developed a three-part framework: trend data enhancement, spatiotemporal feature encoding, and an LLM module.
- Utilized a large language model for capturing generic trends and specific fluctuations.
Main Results:
- Validated LLM-UTP on 11 large-scale datasets across 29 cities/areas.
- Demonstrated superior complexity, scaling law, scalability, generality, and predictive performance.
- Showcased effectiveness across diverse transport modes, traffic scenarios, and time granularities.
Conclusions:
- LLM-UTP shows significant practical potential as a foundation model for intelligent traffic management.
- The framework is suitable for decision-making in future smart cities.
- Highlights the capability of LLMs in addressing complex urban traffic prediction challenges.
