The field of human interaction modeling is moving towards more nuanced and complex representations of social behaviors, incorporating multimodal inputs such as speech, text, and vision. Researchers are developing new frameworks and models to understand and generate human interactions in various contexts, including co-speech gestures, multi-person interactions, and crowd behaviors. These advancements have the potential to improve applications such as human-computer interaction, social robotics, and virtual reality. Notable papers include:

• SocialGen, which introduces a unified motion-language model for modeling multi-human social interactions, and
• Gen-C, which presents a generative model for populating virtual worlds with dynamic crowd behaviors. These works demonstrate significant progress in modeling complex social interactions and have the potential to enable more realistic and engaging virtual environments.