{ "description": "Example benchmark fixture for QMD eval-docs. Tests exact keyword, semantic, and cross-domain retrieval across 6 documents.", "version": 1, "collection": "eval-docs", "queries": [ { "id": "exact-api", "query": "API versioning", "type": "exact", "description": "Direct keyword match in API design document", "expected_files": ["api-design-principles.md"], "expected_in_top_k": 1 }, { "id": "exact-fundraising", "query": "Series A fundraising", "type": "exact", "description": "Direct keyword match in fundraising memo", "expected_files": ["startup-fundraising-memo.md"], "expected_in_top_k": 1 }, { "id": "exact-cap", "query": "CAP theorem", "type": "exact", "description": "Direct keyword match in distributed systems doc", "expected_files": ["distributed-systems-overview.md"], "expected_in_top_k": 1 }, { "id": "semantic-rest", "query": "how to structure REST endpoints", "type": "semantic", "description": "Conceptual match — no exact keyword overlap with 'API design'", "expected_files": ["api-design-principles.md"], "expected_in_top_k": 3 }, { "id": "semantic-fundraising", "query": "raising money for startup", "type": "semantic", "description": "Synonym match — 'raising money' should find 'fundraising'", "expected_files": ["startup-fundraising-memo.md"], "expected_in_top_k": 3 }, { "id": "semantic-overfitting", "query": "how to prevent models from memorizing data", "type": "semantic", "description": "Conceptual match for overfitting in ML primer", "expected_files": ["machine-learning-primer.md"], "expected_in_top_k": 3 }, { "id": "topical-launch", "query": "what went wrong with the product launch", "type": "topical", "description": "Should find the retrospective document", "expected_files": ["product-launch-retrospective.md"], "expected_in_top_k": 3 }, { "id": "cross-domain-consistency", "query": "consistency vs availability tradeoffs", "type": "cross-domain", "description": "CAP theorem concept — specific detail in longer document", "expected_files": ["distributed-systems-overview.md"], "expected_in_top_k": 3 }, { "id": "alias-remote", "query": "working from home guidelines", "type": "alias", "description": "Synonym match — 'working from home' should find 'remote work policy'", "expected_files": ["remote-work-policy.md"], "expected_in_top_k": 3 }, { "id": "hard-partial", "query": "nouns not verbs", "type": "semantic", "description": "Partial phrase recall — API design principle about resource naming", "expected_files": ["api-design-principles.md"], "expected_in_top_k": 5 } ] }