Current State
Only keyword mode has been benchmarked (88.0% R@5). Semantic and hybrid modes require an embedding provider.
Action
Run the benchmark with:
- `BENCHMARK_MODES=semantic EMBEDDING_PROVIDER=ollama`
- `BENCHMARK_MODES=hybrid EMBEDDING_PROVIDER=ollama`
Compare against keyword baseline. Document in RESULTS.md.
Expected Improvement Areas
- single-session-preference (implicit language → semantic similarity)
- temporal-reasoning (date relationships → embedding proximity)
Current State
Only keyword mode has been benchmarked (88.0% R@5). Semantic and hybrid modes require an embedding provider.
Action
Run the benchmark with:
Compare against keyword baseline. Document in RESULTS.md.
Expected Improvement Areas