The field of human mobility and behavior prediction is rapidly advancing, driven by the integration of large language models (LLMs) and innovative machine learning techniques. Researchers are focusing on developing more accurate and efficient models that can capture complex interactions and patterns in human behavior, such as pedestrian trajectory prediction and service-level mobile traffic prediction. The use of LLMs is enabling the creation of more contextual and adaptive models that can better understand and predict human behavior in various environments. Notably, the combination of LLMs with techniques like diffusion models and graph convolutional networks is leading to significant improvements in prediction accuracy and scalability. Some particularly noteworthy papers in this area include:

• InSyn, which proposes a novel Transformer-based model for pedestrian trajectory prediction that explicitly captures diverse interaction patterns.
• DailyLLM, which introduces a lightweight LLM-based framework for context-aware activity log generation that integrates contextual activity information across four dimensions.
• UrbanPulse, which presents a scalable deep learning framework for ultra-fine-grained population transfer prediction that combines a temporal graph convolutional encoder with a transformer-based decoder.