UrbanDS: Graph-Guided LLM Multi-Agent System for Urban Data Tasks
A team of researchers has introduced UrbanDS, a multi-agent system guided by graphs and tailored for data-heavy urban applications. Urban data is characterized by its large scale, diverse sources, and intricate spatial, temporal, and semantic connections. Current LLM agents often depend on restricted datasets and find it difficult to manage such varied data. UrbanDS creates a cohesive dataset graph to streamline the organization of reusable dataset skills and their interconnections. A Data Profiling Agent develops a skill for each dataset, while a Relation Agent uncovers relationships among datasets. This system seeks to automate urban data science tasks, tackling issues of scale and diversity. The research paper can be found on arXiv with ID 2607.26724.
Key facts
- UrbanDS is a graph-guided LLM multi-agent system for data-intensive urban tasks.
- Urban data is large-scale, multi-sourced, and has complex spatial, temporal, and semantic relationships.
- Existing LLM agents face challenges with heterogeneous data repositories.
- A unified dataset graph organizes reusable dataset skills and relationships.
- A Data Profiling Agent constructs a skill for each dataset.
- A Relation Agent identifies relationships among datasets.
- The system automates data science tasks in urban contexts.
- The paper is on arXiv with ID 2607.26724.
Entities
Institutions
- arXiv