Recent advancements in Human-Centric Foundation Models (HFM) have revolutionized the understanding and synthesis of human-related data across perception, generation, and embodied intelligence. These models, trained on vast multimodal datasets, achieve robust perception of human-centric information and high-fidelity synthesis of digital humans. However, despite these successes, current HFM remain fundamentally passive: trained on static datasets with limited support for real-time interaction, they struggle to adapt to real-time human feedback or environmental adaptation. These limitations hinder their deployment in real-world applications in video generation, game industry, and robotics.

This challenge highlights the need for Interactive Human-Centric Foundation Models (I-HFM), a new research direction in the AI community, designed to perceive, generate, and act in ways that seamlessly align with human needs and behaviors through bidirectional engagement between humans, environments, and AI systems. I-HFMs cover three critical dimensions of interactivity:
(a) Interact with users, where HFM empower users to intuitively guide the creation and refinement of human-centric content, (e.g., personalized avatars, animations, or virtual try-ons), with natural language, gestures, or iterative feedback.
(b) Interact with physical or virtual environments, enabling I-HFMS to learn skills from environmental feedback, like humans, by integrating perception, reasoning, and action to adapt to dynamic scenarios.
(c) Interact with other agents (including humans) to collaborate to solve complex, socially embedded tasks in the real world.

This interactivity of HFM is not merely a technical advancement feature but a fundamental step toward achieving AGI. By enabling continuous adaptation to human feedback and real-world dynamics, I-HFMs transform AI from passive perception models into proactive, human-like autonomous agents.This ability to perceive, reason, and act in environments as human beings bridges the gap between static data-driven systems and truly responsive and socially intelligent AI, paving the way for AGI that seamlessly integrates into human societies.