Scaling Tasks, Not Samples: Mastering Humanoid Control through Multi-Task Model-Based Reinforcement Learning

Researchers propose EfficientZero-Multitask (EZ-M), a multi-task model-based reinforcement learning algorithm that scales the number of tasks rather than samples per task for robotics training. The approach achieves state-of-the-art performance on HumanoidBench with significantly higher sample efficiency by leveraging shared world models across diverse tasks.

• →EZ-M demonstrates that scaling tasks rather than samples leads to more efficient robotic learning through shared world models.
• →Model-based reinforcement learning shows structural advantages over model-free methods in multi-task scenarios due to invariant physical dynamics.
• →Task diversity acts as a natural regularizer for model-based RL, improving dynamics learning and sample efficiency.
• →The approach achieves state-of-the-art results on HumanoidBench without requiring extreme parameter scaling.
• →Multi-task experience aggregation enables learning of robust, task-agnostic representations for embodied AI systems.