The competition aims to address real-world generalization challenges for embodied AI foundation models, fostering advances in single-model, general-purpose multi-task capabilities.