{"record": {"aliases": ["Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA"], "artifacts": [{"id": "arxiv:2605.29277", "kind": "paper", "url": "https://arxiv.org/abs/2605.29277"}], "categories": ["Code & Software", "Reasoning", "Software & AI Compute"], "description": {"en": "An open-source framework for generating repository-level question-answering benchmarks under closed-book, code-only, and documented conditions."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.29277", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.29277", "title": "Code-QA-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.29277", "modality": null, "name": "Code-QA-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_code-qa-bench-separating-code-reasoning-fr_045c6a06", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "b6ac06635a70c597a3228f07bd921bdee1d280c11bf5c26e71aa682d6a8d8ffe", "relation": "evaluates_only", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T19:45:52.845491Z", "source_url": "https://arxiv.org/abs/2605.29277"}, "publisher": null, "released": "2026-05-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.29277", "source_url": "https://arxiv.org/abs/2605.29277"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-29277", "source": "claire_radar", "source_benchmark_id": "2605.29277", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA"], "annotation": "Unknown", "applicationDomains": ["Software & AI Compute"], "area": "Code & Software", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Repository-level code evaluation and memorization controls are broadly relevant, and the framework is described as open-source.", "score": 63}, "capabilities": ["Reasoning"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T19:45:52.845491Z", "inputHash": "ec84cad8170ab5c0937b2ec3a2098bcbb74be300da3cdf5cf81d43776ba61b5f", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA", "canonicalNameSource": "paper_title", "decisionReason": "The source supplies a formal name, a repeatable three-condition task design, named judging dimensions, and an explicit current open-source reuse statement.", "evaluationEvidence": {"reuse": "The framework is open-source and applicable to any well-documented Python repository.", "scoring": "scored by an LLM judge on accuracy, completeness, and specificity", "task": "repository-level code understanding benchmarks that separates genuine code comprehension from documentation recall and pretraining memorization"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T19:45:52.845491Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "An open-source framework for generating repository-level question-answering benchmarks under closed-book, code-only, and documented conditions.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We present Code-QA-Bench, a fully automated framework for synthesizing repository-level code understanding benchmarks that separates genuine code comprehension from documentation recall and pretraining memorization."}, "familyId": "bmf_022ea985594d", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_code-qa-bench-separating-code-reasoning-fr_045c6a06", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Software & Cloud"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2605.29277", "pdf": "https://arxiv.org/pdf/2605.29277", "project": null, "report": "https://arxiv.org/abs/2605.29277"}, "metrics": [], "motivation": "We present Code-QA-Bench, a fully automated framework for synthesizing repository-level code understanding benchmarks that separates genuine code comprehension from documentation recall and pretraining memorization.", "name": "Code-QA-Bench", "oneLine": "An open-source framework for generating repository-level question-answering benchmarks under closed-book, code-only, and documented conditions.", "paperTitle": "Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA", "primaryDomain": "Software & AI Compute", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.29277", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-05-28", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Jun Zhang", "JianYing Qu", "Hanwen Du", "Zhongkai Sun", "Yehua Yang", "Qiao Zhao"], "categories": ["cs.SE", "cs.AI"], "id": "2605.29277", "title": "Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA", "type": "arxiv", "url": "https://arxiv.org/abs/2605.29277"}, "sourceUpdatedAt": "2026-05-28T02:52:58Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "It separates code comprehension from documentation utility and possible memorization through controlled evaluation conditions."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
