LLM-Driven SWMM Scenario Simulation Agent Data for Lili Town, Suzhou

1.1 MB of data and code from experiments on an LLM-driven agent for the Storm Water Management Model (SWMM). The agent integrates geospatial awareness to automate parameter adjustment for urban flood simulation, tested in the Lili Town, Suzhou, China study area. The dataset, authored by Yani Zhong and last updated in April 2026, includes JSON and ZIP files.

Use Cases

• Benchmarking LLM performance on geospatial reasoning tasks for hydrological models based on the described agent modules.
• Training or validating agents for automated SWMM parameter adjustment based on natural language instructions.
• Studying the integration of hierarchical geographic entity disambiguation and spatial reasoning in environmental simulation workflows.

Strengths

• Includes experimental results showing a 78.57% overall success rate for the agent.
• Data supports testing across seven mainstream LLMs, with performance metrics for Kimi-k2.5 and Claude-Sonnet-4.6 provided.
• Released under a permissive CC-BY-4.0 license for open reuse.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The 1.1 MB size suggests a small-scale experimental dataset.

Provenance

Source

figshare, authored by Yani Zhong.

Collection Method

Likely contains experimental data and code from testing an LLM-driven agent for SWMM simulation.

Freshness

Last updated 2026-04-21 01:51:08; freshness should be verified.

Geography

Study area around Lili Town, Suzhou, China.