The structure and distribution of reward signals across different tasks, which can vary significantly in multimodal learning.