用大模型生成虚拟世界中的智能人群，让群体行为更自然真实。

Over the past two decades, researchers have made significant steps in simulating agent-based human crowds, yet most efforts remain focused on low-level tasks such as collision avoidance, path following, and flocking. As a result, these approaches often struggle to capture the high-level behaviors that emerge from sustained agent-agent and agent-environment interactions over time. We introduce Generative Crowds (Gen-C), a generative framework that produces crowd scenarios capturing agent-agent and agent-environment interactions, shaping coherent high-level crowd plans. To avoid the labor-intensive process of collecting and annotating real crowd video data, we leverage Large Language Models (LLMs) to bootstrap synthetic datasets of crowd scenarios. To represent those scenarios, we propose a time-expanded graph structure encoding actions, interactions, and spatial context. Gen-C employs a dual Variational Graph Autoencoder (VGAE) architecture that jointly learns connectivity patterns and node features conditioned on textual and structural signals, overcoming the limitations of direct LLM generation to enable scalable, environment-aware multi-agent crowd simulations. We demonstrate the effectiveness of our framework on scenarios with diverse behaviors such as a University Campus and a Train Station, showing that it generates heterogeneous crowds, coherent interactions, and high-level decision-making patterns consistent with the provided context.