TechiSeek

VendorsFor Vendors

Encounter-Driven City-Scale Human Mobility Generation with Large Language Models

arXiv cs.AI

Encounter-Driven City-Scale Human Mobility Generation with Large Language Models

Encounter Mob is an encounter-driven LLM framework for city-scale human mobility generation that jointly preserves individual mobility and population-level interaction structure.

Examples

  • Experiments on the Japan and NYC datasets: Encounter Mob outperforms all external baselines on all seven encounter metrics and all six personal mobility metrics.
  • Compared with ELLMob, it reduces the participant-time-venue joint error by 40.4% on Japan and 41.7% on NYC.
Read More