-
Release v2.0 with provenance and quality audits
Quality-gated deterministic release: row-level provenance, entity/template-disjoint splits, full exact deduplication, split leakage checks, pinned benchmark contamination audit, datasheet, and reproducible generation/audit code. -
Scale dataset to 500,000 validated rows in sharded Parquet
Replace the small JSONL release with deterministic train/validation/test Parquet shards, a transparent generation card, and verified split counts. -
Publish Turkish synthetic dataset with train/test splits