Collect user feedback on LLM or agent outputs to improve model performance and user satisfaction.