- Add annotations to explain what the confusion matrix, accuracy, and F1-score represent - Include legends or text labels that highlight what a "good" or "poor" result looks like - Possibly summarize model performance with one-sentence captions under each chart - Add a brief text-based summary of how each model performed