TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation
Researchers have introduced TransitLM, a novel dataset and benchmark designed for map-free public transit route generation. This large-scale dataset, comprising over 13 million transit records from four Chinese cities, enables large language models to plan routes without relying on traditional map infrastructure. Experiments demonstrate that LLMs trained on TransitLM can accurately generate valid routes and implicitly map GPS coordinates to stations, showcasing the potential for end-to-end, data-driven route planning. AI
IMPACT Enables LLMs to perform complex route planning tasks without traditional map data, potentially improving navigation and logistics applications.