The field of city sciences is moving towards more realistic simulations of human behavior in urban environments, with a focus on developing models that can accurately predict travel mode choices and generate realistic traffic scenarios. This is being achieved through the integration of Large Language Models (LLMs) with other techniques such as Graph Retrieval-Augmented Generation (RAG) and diffusion models. These models are able to capture complex human behaviors and generate realistic outputs, making them useful for applications such as urban mobility modeling and self-driving system testing. Noteworthy papers in this area include: The paper introducing the Preference Chain method, which enhances context-aware simulation of human behavior in transportation systems. The paper proposing a diffusion-based multi-head trajectory planner, which enables dynamic instruction-aware planning without switching models. The paper introducing HDSim, an HD traffic generation framework that combines cognitive theory with LLM assistance to produce scalable and realistic traffic scenarios. The paper evaluating Retrieval-Augmented Generation strategies for LLMs in travel mode choice prediction, which demonstrates the importance of aligning retrieval strategies with model capabilities.