Mahdi Bazrafshan d572724115 docs(data): add 21-question evaluation dataset for benchmarking ...

Why:
- Need standardized questions to evaluate chunking strategies
- Questions cover easy/medium/hard difficulty levels
- Golden answers enable automated scoring

Changes:
- 21 questions about engineering insurance (بیمه‌های مهندسی)
- 7 easy, 7 medium, 7 hard questions
- Categories: factual, procedural, comparative, complex
- Each question includes golden answer and expected keywords

2026-07-26 13:48:36 +03:30

15 KiB

Raw Blame History

View Raw