Zijie Zhao

We propose a novel framework employs a MoE-based multi-agent world model for learning and planning in multi-task scenarios.