corpus: codebase-index (../codebase-index) � 87 queries from research/data/codebase-index.yml
corpus: Civitas (../Civitas) � 64 queries from research/data/Civitas.yml
corpus: PoliternalSite (../PoliternalSite) � 120 queries from research/data/PoliternalSite.yml
corpus: PoliternalParkour (../PoliternalParkour) � 64 queries from research/data/PoliternalParkour.yml
corpus: TerraForge (../TerraForge) � 17 queries from research/data/TerraForge.yml
corpus: denfry.github.io (../denfry.github.io) � 39 queries from research/data/denfry.github.io.yml
corpus: DevGraph (../DevGraph) � 12 queries from research/data/DevGraph.yml
corpus: Windows-Cleaner-and-Optimizer-main (../Windows-Cleaner-and-Optimizer-main) � 17 queries from research/data/WinCleaner.yml
pooled: 420 queries across 8 corpora
building index for codebase-index (once, shared by all variants)...
building index for Civitas (once, shared by all variants)...
building index for PoliternalSite (once, shared by all variants)...
building index for PoliternalParkour (once, shared by all variants)...
building index for TerraForge (once, shared by all variants)...
building index for denfry.github.io (once, shared by all variants)...
building index for DevGraph (once, shared by all variants)...
building index for Windows-Cleaner-and-Optimizer-main (once, shared by all variants)...

| label | recall@5 | recall@10 | MRR | oracle | eff | nDCG@10 | hit@3 | P@5 | MAP | useful@budget | tokens | dup% | p50_ms | p95_ms |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| baseline (1.7.0) | 0.376 | 0.481 | 0.318 | 0.714 | 0.446 | 0.329 | 0.383 | 0.130 | 0.256 | 0.468 | 508.1 | 1.3 | 24.2 | 40.1 |
| previous release (1.9.0) | 0.589 (+0.214) | 0.623 (+0.142) | 0.560 (+0.242) | 0.900 (+0.186) | 0.622 (+0.177) | 0.522 (+0.193) | 0.655 (+0.271) | 0.190 (+0.059) | 0.452 (+0.196) | 0.583 (+0.115) | 1099.4 (+591.3) | 0.0 (-1.3) | 48.6 (+24.4) | 95.0 (+54.9) |
| default (all signals) | 0.610 (+0.235) | 0.687 (+0.205) | 0.580 (+0.262) | 0.900 (+0.186) | 0.645 (+0.199) | 0.554 (+0.225) | 0.667 (+0.283) | 0.197 (+0.067) | 0.475 (+0.219) | 0.613 (+0.144) | 1083.6 (+575.5) | 0.0 (-1.3) | 53.0 (+28.9) | 102.3 (+62.2) |

paired comparison: default (all signals) vs previous release (1.9.0) (n=420)
| metric        |    delta |              95% CI |      p | sig |
|---|---|---|---|---|
| recall@5      |  +0.0210 | [+0.0111,+0.0329] |  0.000 | yes |
| recall@10     |  +0.0631 | [+0.0429,+0.0861] |  0.000 | yes |
| MRR           |  +0.0200 | [+0.0076,+0.0326] |  0.002 | yes |
| nDCG@10       |  +0.0316 | [+0.0213,+0.0425] |  0.000 | yes |
| hit@3         |  +0.0119 | [+0.0000,+0.0262] |  0.175 | no  |
| P@5           |  +0.0075 | [+0.0043,+0.0113] |  0.000 | yes |
| MAP           |  +0.0230 | [+0.0127,+0.0348] |  0.000 | yes |
| useful@budget |  +0.0296 | [+0.0056,+0.0538] |  0.011 | yes |
| oracle        |  +0.0000 | [+0.0000,+0.0000] |  1.000 | no  |
| cand_recall   |  +0.0000 | [+0.0000,+0.0000] |  1.000 | no  |

Per-corpus (default config):
| label | recall@5 | recall@10 | MRR | oracle | eff | nDCG@10 | hit@3 | P@5 | MAP | useful@budget | tokens | dup% | p50_ms | p95_ms |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| codebase-index | 0.548 | 0.634 | 0.649 | 0.931 | 0.697 | 0.552 | 0.736 | 0.226 | 0.459 | 0.617 | 1193.3 | 0.0 | 53.1 | 80.7 |
| Civitas | 0.546 | 0.633 | 0.572 | 0.938 | 0.610 | 0.529 | 0.656 | 0.175 | 0.462 | 0.548 | 921.5 | 0.0 | 55.0 | 137.8 |
| PoliternalSite | 0.560 | 0.631 | 0.548 | 0.875 | 0.626 | 0.516 | 0.625 | 0.184 | 0.441 | 0.570 | 1101.8 | 0.0 | 72.1 | 124.8 |
| PoliternalParkour | 0.695 | 0.809 | 0.488 | 0.875 | 0.557 | 0.561 | 0.609 | 0.147 | 0.479 | 0.680 | 931.9 | 0.0 | 47.8 | 112.0 |
| TerraForge | 0.706 | 0.740 | 0.669 | 1.000 | 0.669 | 0.646 | 0.824 | 0.282 | 0.571 | 0.549 | 933.8 | 0.0 | 43.3 | 114.9 |
| denfry.github.io | 0.656 | 0.720 | 0.577 | 0.795 | 0.725 | 0.574 | 0.667 | 0.191 | 0.505 | 0.720 | 1226.9 | 0.0 | 22.3 | 43.8 |
| DevGraph | 0.819 | 0.819 | 0.694 | 1.000 | 0.694 | 0.667 | 0.750 | 0.317 | 0.574 | 0.778 | 1260.8 | 0.0 | 33.6 | 52.7 |
| Windows-Cleaner-and-Optimizer-main | 0.863 | 0.863 | 0.675 | 0.941 | 0.718 | 0.677 | 0.647 | 0.261 | 0.583 | 0.578 | 1270.2 | 0.0 | 24.7 | 40.4 |

Per-category MRR (default config):
  bug            0.647
  change         0.529
  docs           0.576
  feature        0.643
  perf           1.000
  refactor       0.750
  test           0.494
