retrieval-eval
Steps
- Build 50+ question/relevant-chunk pairs.
- Measure recall@k and MRR.
- Measure answer faithfulness separately from retrieval.
- Track results per change in a table.
Done When
- [ ] Baseline recorded before any change
Registry
| id | role | version | status | owner | last_tested |
|---|---|---|---|---|---|
| SK-RAG-ENGINEER-retrieval-eval | rag-engineer | 0.1.0 | seed | Thomas Cherickal | - |
Links
- Up: RAG Engineer skills
- Siblings: chunking-strategy, hybrid-search-rerank
- Frameworks: react, few-shot, risen, xml-structured, chain-of-thought
- Tests: testing guide