Stochastic Streets: A Walk Through Random LLM Address Generation in four European Cities

2025-09-16

AI Summary

This work presents the first systematic evaluation of large language models (LLMs) on city-scale stochastic street address generation in European urban contexts. Focusing on Berlin, Paris, Rome, and Warsaw, the study quantitatively assesses generated addresses along three dimensions: geographic accuracy, syntactic validity, and statistical randomness—using real-world address datasets as ground truth. Methodologically, it integrates prompt engineering, pattern-based reasoning, and empirical benchmarking, deliberately avoiding hand-crafted rules. Results show that while LLMs produce syntactically correct addresses partially aligned with empirical distributions, they exhibit substantial geographic inaccuracies—including fabricated postal codes and cross-district street–district mismatches—as well as spurious randomness, manifesting as repetitive structural patterns. The study delineates the implicit limits of LLMs’ internal modeling of structured geospatial knowledge and establishes a reproducible, multi-dimensional evaluation framework for geospatial text generation. It further provides a critical baseline for advancing trustworthy LLM deployment in location intelligence applications.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large Language Models (LLMs) are capable of solving complex math problems or answer difficult questions on almost any topic, but can they generate random street addresses for European cities?

Problem

• Generating random street addresses using LLMs
• Testing LLM capability for European cities
• Evaluating accuracy in address generation

Innovation

• Random street address generation
• Using Large Language Models
• For four European cities