Google Research Introduces ME-POIs: A Mobility-Informed Framework that Adds "How a Place Is Used" to Text-Based POI Embeddings

Google Research and USC released ME-POIs, a framework that adds aggregate human-movement signals to text-based place embeddings, and reports it improved 34 of 35 model-task pairings on Los Angeles map-enrichment benchmarks. The model uses Space2Vec and Time2Vec encoders with contrastive learning, reaching gains up to 81.9% F1 on visit intent. A mobility-only variant beat Gemini embeddings on price-level classification. No public code or weights are available yet.