{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.22061", "kind": "paper", "url": "https://arxiv.org/abs/2606.22061"}], "authors": ["Heejin Do", "Alexandre Ballenghien", "Yang Wu", "April Yi Wang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.22061", "source_url": "https://arxiv.org/abs/2606.22061"}, "categories": ["Code & Software", "Code generation", "General AI", "cs.CL"], "description": {"en": "NL2Scratch is an executable benchmark for natural-language-to-Scratch generation, consisting of 311,648 parser-valid NL-program pairs extracted from real Scratch projects. It includes a semantically validated pool of 23,594 examples and an 800-example diagnostic set. Evaluation uses Semantic Alignment Consistency (SAC), an interpretable slot-level metric for measuring semantic agreement."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.22061", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.22061", "title": "NL2Scratch"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["ETH Zurich"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.22061", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.22061", "source_url": "https://openalex.org/W7165621433"}, "key": "claire-radar:2606.22061", "modality": null, "name": "NL2Scratch", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_nl2scratch_fe2ce42e", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "69a72775fbf94e4d649ae7002312bbfce5307e107b9e8072e49b4f02c6651846", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.22061"}, "publisher": null, "released": "2026-06-20", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.22061", "source_url": "https://arxiv.org/abs/2606.22061"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-22061", "source": "claire_radar", "source_benchmark_id": "2606.22061", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Code & Software", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.22061", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": ["Code generation"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "5f22e85182740424ca0a409b023042bb9a0624de75b20a79512b1d1dfb8e8c59", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "NL2Scratch is an executable benchmark for natural-language-to-Scratch generation, consisting of 311,648 parser-valid NL-program pairs extracted from real Scratch projects. It includes a semantically validated pool of 23,594 examples and an 800-example diagnostic set. Evaluation uses Semantic Alignment Consistency (SAC), an interpretable slot-level metric for measuring semantic agreement.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce NL2Scratch, an executable benchmark for natural-language-to-Scratch generation comprising 311,648 parser-valid NL--program pairs, whose program side is extracted from real Scratch projects and paired with semantically aligned NL descriptions."}, "familyId": "bmf_3590e32806d4", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_nl2scratch_fe2ce42e", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.22061", "paper": "https://arxiv.org/abs/2606.22061", "pdf": "https://arxiv.org/pdf/2606.22061", "project": null, "report": "https://arxiv.org/abs/2606.22061"}, "metrics": [], "motivation": "Block-based programming environments such as Scratch are widely used in early programming education, yet natural-language-to-code (NL2Code) research has focused primarily on text-based languages.", "name": "NL2Scratch", "oneLine": "NL2Scratch is an executable benchmark for natural-language-to-Scratch generation, consisting of 311,648 parser-valid NL-program pairs extracted from real Scratch projects. It includes a semantically validated pool of 23,594 examples and an 800-example diagnostic set. Evaluation uses Semantic Alignment Consistency (SAC), an interpretable slot-level metric for measuring semantic agreement.", "paperTitle": "NL2Scratch: An Executable Benchmark and Evaluation for Block-Based Programming", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.22061", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "NL2Scratch Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://arxiv.org/abs/2606.22061"}], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-20", "source": {"authors": ["Heejin Do", "Alexandre Ballenghien", "Yang Wu", "April Yi Wang"], "categories": ["cs.CL", "cs.AI"], "id": "2606.22061", "title": "NL2Scratch: An Executable Benchmark and Evaluation for Block-Based Programming", "type": "arxiv", "url": "https://arxiv.org/abs/2606.22061"}, "sourceUpdatedAt": "2026-06-20T14:22:59Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "Existing NL2Code evaluation focuses on text-based languages, leaving block-based programming unevaluated. NL2Scratch enables assessment of models on event-driven, visually compositional programs, revealing gaps between lexical similarity and semantic alignment that are invisible under token-level metrics."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
