Workload
rag-pipeline-ru-v0
Internal
Question it answers
Where does a Russian-language RAG pipeline lose quality, component by component?
What is measured
Per-component evaluation: OCR CER/WER → splitter boundary F1 → Recall@k/nDCG → reranker delta → generation support/abstention/citations. Internal until 6 release gates pass.
Claims this workload cannot support
Any commercial use before release gates; end-to-end quality claims from a single component metric.
Identity (not frozen yet)
This workload is not released: its corpora, hashes and acceptance checks are still pending, and it backs no public headline claims. The declared contract below is read from the catalog file, not typed into this page.
- Revision
- 2026-07-28-draft
- Load model
- mixed
- Quality gates
- six_release_gates_pending
- Metrics contract
- ocr_cer_wer, splitter_boundary_f1, recall_at_k, ndcg, reranker_delta, support_abstention_citations
- Corpus hash
- not frozen yet