arborist/bench/qa_questions_brit.txt
russell@unturf.com b573c592d8
feat(retrieval): accent-fold (+30pp recall@1) + fold-search factory hardening
Second MEASURED fold-search win, and the instrument correcting my own
premature call. accent-fold ON vs OFF on the mined accent fixture:
recall@1 55% -> 85% (+30pp), rank-1 22/40 -> 34/40. recall@8 was
flat (95->98) — a too-lenient k nearly got a real lever wrongly
reverted; @1/@3 is the resolution that drives primary-source
selection. _accent_fold_variants: ASCII-fold then re-tokenise so a
diacritic title ("Béla Bartók", which _TITLE_TOKEN_RE otherwise
fragments to junk) matches the ASCII form a user types. Additive+
symmetric, no-op on pure-ASCII (zero effect on non-accent
queries/titles), mirrors _hyphen_fold_variants (#000007).

Also fixes a defect I shipped in a3ac653: an orphaned duplicate
body left as dead code after `return base` in _title_query_tokens
(unreachable — numeral-fold behaviour/measurement were valid — but
cruft; removed).

Fold-search factory, fanned out across the full survey backlog
(deterministic recall, no LLM, parallel — serial-by-caution was
halting in disguise):
- recall_at_k.py: returns rank -> recall@1/@3/@k from one retrieval
  (verified offline). A coarse k hides rank-only lifts.
- mine_questions.py: numeral/accent/hyphen/honorific/amp/brit
  ground-truth classes; fixtures committed.
- Measured @1 headroom verdicts: accent SHIP (this commit);
  honorific 45% / brit 50% = real headroom (build next); hyphen
  90% = existing #000007 already delivers, NOTHING to build (the
  measure-the-unmeasured-thing check pays off); amp 82% = no fold
  needed (prevalence-overranked, instrument kills it cheaply).

CLAUDE.md bench-maxing: two measured lessons codified — report
recall@1/@3/@k (a lenient k hides rank lifts; prevalence != miss-
rate), and fan out independent measurements (serial-by-caution is
halting). Full suite 2488 passed, 0 regressions (accent-fold is
hot-path in _title_query_tokens); real-path test (FakeSource->
ingest->query()->real _Hit).
2026-05-18 19:23:22 -04:00

43 lines
1.7 KiB
Text

# AUTO-MINED (brit class) from corpus titles via bench/mine_questions.py — ground-truth-carrying.
# Graded by deterministic retrieval recall@k (bench/recall_at_k.py), NOT audit_mode. Not adversarial; complements (never replaces) qa_questions.txt.
what is Finnish defense Forces?
what is Hopewell center, Hong Kong?
what is labor economics?
what is World organization for Animal Health?
what is Papua New Guinea defense Force?
what is programr?
what is Tonga defense Services?
what is Trinidad and Tobago defense Force?
what is Zambian defense Force?
what is defense Signals Directorate?
what is English Renaissance theater?
what is labor Party (Netherlands)?
what is The Luzhin defense?
what is Asia-Pacific Network Information center?
what is Soyuz program?
what is center Georges Pompidou?
what is Amphitheater?
what is Australian defense Force?
what is center County, Pennsylvania?
what is organization internationale de la Francophonie?
what is Luna program?
what is Man and the Biosphere program?
what is centerville, Maryland?
what is Boggs Township, center County, Pennsylvania?
what is Ferguson Township, center County, Pennsylvania?
what is Harris Township, center County, Pennsylvania?
what is Huston Township, center County, Pennsylvania?
what is Patton Township, center County, Pennsylvania?
what is Union Township, center County, Pennsylvania?
what is Worth Township, center County, Pennsylvania?
what is center Township, Perry County, Pennsylvania?
what is center Township, Pennsylvania?
what is Tricorn center?
what is Living color?
what is Philipsburg, center County, Pennsylvania?
what is German labor Front?
what is Rogers center?
what is Ulster defense Association?
what is Guiana Space center?
what is Community theater?