miroir/tests
jedarden 0de5f01d32 P2.2: Pluggable MergeStrategy trait + RRF scoring + full benchmark re-run
- Extract MergeStrategy trait with merge()/name() methods
- Implement RrfStrategy with configurable k (default 60)
- Refactor scatter_gather_search to accept &dyn MergeStrategy
- Add RRF simulation to benchmark script (simulate_distributed_search_rrf)
- Re-run full benchmark (3989 queries) with updated comparison reports
- Add topology unit tests (NodeId, NodeStatus, Node helpers)

Benchmark results:
  Score-based merge: avg tau = 0.798 (FAIL, common-term tau = 0.152)
  RRF merge:         avg tau = 0.134 (FAIL, rank-only loses score signal)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-04-19 02:07:39 -04:00
..
benches/score-comparability P2.2: Pluggable MergeStrategy trait + RRF scoring + full benchmark re-run 2026-04-19 02:07:39 -04:00