UrbanDS: A Graph-Guided LLM Multi-Agent System for Data-Intensive Urban Tasks
- Source
- arxiv.org
- Author
- Zhilun Zhou, Jianghao Yu, Yuming Lin, yongjun yang, Sun Yongquan, Depeng Jin, Yong Li
- Date
Context
UrbanDS addresses a step that comes before urban data analysis: finding which datasets belong together. It organizes datasets and their spatial, temporal, or semantic relationships into a graph. A planning uses those connections to select relevant data and arrange the work, while execution agents share intermediate results and a reporting agent assembles the findings. System description.
The authors report stronger performance than comparison data-science agents on their evaluations and deployment on an urban operations platform in Wuhan's Dongxihu District. The supported contribution is a workflow for discovering and combining information across a large repository. The deployment statement establishes reported use, without quantifying improvements in city services. Study abstract.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
Checking sign-in…
Loading comments…