MedLFQA

MedLFQA is reconstructed by reformulating the current four biomedical long-form question-answering benchmark datasets: LiveQA, MedicationQA, HealthsearchQA, and K-QA. MedLFQA consists of four components: question (Q), answer (A), must-have statements (MH), and nice-to-have statements (NH). It facilitates the automatic evaluation of models' responses and provides a comprehensive understanding of how the model responds to a patient's question.