The field of embodied intelligence is rapidly advancing, with a strong focus on developing world models that enable agents to perceive, reason, and act within their environments. Recent research has highlighted the importance of integrating multimodal perception, planning, and memory to create comprehensive world models. Moreover, the development of embodied social agents with lifelong memory systems has shown great promise in advancing autonomous decision-making and social interaction capabilities. The use of physical simulators and world models has emerged as a key enabler in this quest, allowing for the development of more generalizable and adaptable embodied AI systems. Noteworthy papers in this area include:

• Ella: Embodied Social Agents with Lifelong Memory, which introduced a structured, long-term multimodal memory system for embodied social agents.
• RoboBrain 2.0 Technical Report, which presented a heterogeneous architecture for embodied vision-language foundation models, achieving strong performance across a wide spectrum of embodied reasoning tasks.