Add 5 markdown documents that users can index to reproduce evaluation results.
Changes:
- Add sample_documents/ folder with 5 markdown files covering LightRAG features
- Update sample_dataset.json with 3 improved, specific test questions
- Shorten and correct evaluation README (removed outdated info about mock responses)
- Add sample_documents reference with expected ~95% RAGAS score
Test Results with sample documents:
- Average RAGAS Score: 95.28%
- Faithfulness: 100%, Answer Relevance: 96.67%
- Context Recall: 88.89%, Context Precision: 95.56%
