TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation

TransitLM is a continual pre-training corpus and benchmark of 13M+ transit planning records across 4 Chinese cities, covering 120,845 stations and 13,666 lines, with 3 evaluation tasks. LLMs trained on it produce structurally valid routes without map infrastructure or routing engines. Relevant to transit AI, urban mobility, and teams building map-free navigation systems.