Extracts questions from the 8 raw questionnaires and compares against human-cleaned ground truth. Reports precision, recall, and F1.
Click “Run Benchmarks” to evaluate extraction quality across all 8 questionnaire files.