AI Attention Budget - Graph View

The finite computational attention a language model distributes across tokens in its context, where quality degrades as the model must spread attention over more content.