A development team is using a neural network model to process legal documents. They observe that processing a 1,000-token document excerpt takes approximately 3 seconds. However, when they attempt to process a full 10,000-token document, the processing time increases to approximately 5 minutes (300 seconds). Based on this performance data, what is the most likely computational characteristic of the model's architecture that explains this disproportionate increase in processing time? Explain your reasoning using the data provided.