According to arxiv.org, researchers have introduced UrbanDS, a graph-guided large language model (LLM) multi-agent system designed to handle data-intensive urban tasks. The system addresses challenges in scenarios requiring discovery and use of relevant information from large-scale, heterogeneous data repositories.
According to the research paper, UrbanDS constructs a unified dataset graph to organize reusable dataset skills and relationships among datasets. The system employs multiple specialized agents: a Data Profiling Agent that constructs a skill for each dataset, a Relation Agent that identifies dataset relationships, a Planner Agent that retrieves task-relevant datasets and generates execution plans, multiple Execution Agents that perform data processing and analysis with shared memory, and a Report Agent that synthesizes experimental logs.
The researchers built UrbanDS-Bench, an urban data science benchmark covering representative data analysis and modeling tasks, according to arxiv.org. Experiments on both general and urban benchmarks demonstrated that UrbanDS “consistently outperforms existing data science agents on data-intensive tasks,” the paper states.
According to the source, UrbanDS has been deployed on the urban operations platform of Dongxihu District in Wuhan, demonstrating effectiveness in real-world urban applications. The research was published on July 31, 2026, on arxiv.org under the classification of Artificial Intelligence (cs.AI).