feat(datasets): Phase 1 in-house sets — domain-bilingual-v1 + negatives-ood-v1#7

Merged

helebest merged 7 commits into

mainfrom

eval/phase1-inhouse-datasets

Jun 28, 2026

Commits on Jun 28, 2026

docs: design for Phase 1 in-house datasets (domain-bilingual-v1 + negatives-ood-v1)

authored and
helebest
committed
docs: implementation plan for Phase 1 in-house datasets

authored and
helebest
committed
feat(tools): per-language metric splitter for bilingual eval NDJSON

authored and
helebest
committed
feat(datasets): scaffold domain-bilingual-v1 + negatives-ood-v1 (corpus copy)

authored and
helebest
committed
feat(datasets): curated bilingual positives + OOD negatives; LLM-gen via factory

authored and
helebest
committed
feat(datasets): calibrate domain-bilingual-v1 thresholds at observed-margin

authored and
helebest
committed
eval: record Phase 1 in-house calibration (domain-bilingual-v1 + negatives-ood-v1)

authored and
helebest
committed