All benchmark evaluation results are available on Google Drive:
The Google Drive folder contains results from various configurations:
- results-base_retrieval/ - Baseline retrieval-only results
- results-best/ - Best performing configuration results
- results-context/ - Context-enhanced results
- results-context_expansion/ - Context + expansion results
- results-expansions/ - LLM expansion results
- results-hierarchy/ - Hierarchy-enhanced results
- results-hierarchy_expansion/ - Hierarchy + expansion results
- results-llm/ - Full LLM pipeline results
- summary-json/ - Summary statistics in JSON format
- code/ - Code used for evaluation
- ALL_CONFIGURATIONS.md - Complete configuration descriptions
Each result folder typically contains:
- Performance metrics (accuracy, precision, recall, F1)
- Per-sample predictions
- Configuration details
- Timing information