{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.17861", "kind": "paper", "url": "https://arxiv.org/abs/2606.17861"}, {"kind": "website", "url": "https://tongxuluo.github.io/gamecraft-bench-website"}], "authors": ["Tongxu Luo", "Rongsheng Wang", "Jiaxi Bi", "Chenming Xu", "Zhengyang Tang", "Jianlong Chen", "Juhao Liang", "Ke Ji", "Shuqi Guo", "Yuhao Du", "Fan Bu", "Wenyu Du", "Xiaotong Zhang", "Kyle Li", "Shaobo Wang", "Linfeng Zhang", "Yuxuan Liu", "Xin Lai", "Chenxin Li", "Yiduo Guo", "Zhexin Zhang", "Xinyuan Wang", "Tianyi Bai", "Ziniu Li", "Benyou Wang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.17861", "source_url": "https://arxiv.org/abs/2606.17861"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "A benchmark of 140 Godot engine tasks across 15 game families. Coding agents must build playable games end-to-end, and the games are scored through replayed demonstrations and rubric-guided multimodal judging."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.17861", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.17861", "title": "GameCraft-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Chinese University of Hong Kong, Shenzhen", "Tencent (China)", "National University of Singapore", "Shanghai Jiao Tong University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.17861", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.17861", "source_url": "https://openalex.org/W7165025980"}, "key": "claire-radar:2606.17861", "modality": null, "name": "GameCraft-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_gamecraft-bench_5127c2aa", "origin_source": "arxiv", "recognition_confidence": "0.75", "record_sha256": "e2d84c31c55fe9c44168a7fef21cb2a01c3ef38d4e62384e8a6ed627b2573d86", "relation": "introduces", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T16:34:01.178138Z", "source_url": "https://arxiv.org/abs/2606.17861"}, "publisher": null, "released": "2026-06-16", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.17861", "source_url": "https://arxiv.org/abs/2606.17861"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-17861", "source": "claire_radar", "source_benchmark_id": "2606.17861", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Agentic game generation in a real engine is a new area. Low scores leave room for progress, and the project site provides demos, code, and data.", "score": 55}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T16:34:01.178138Z", "inputHash": "c8549954dd22e8086bfac826fe7a20b21f8325b8936943991296f29e807f5df4", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We instantiate this framework as GameCraft-Bench", "canonicalNameSource": "abstract", "decisionReason": "A formally named benchmark with a defined task set, a stated scoring protocol, and public code and data linked from the abstract.", "evaluationEvidence": {"reuse": "See https://tongxuluo.github.io/gamecraft-bench-website for demos, code, and data.", "scoring": "assesses executable gameplay through replayed demonstrations and rubric-guided multimodal judging", "task": "a benchmark comprising 140 Godot tasks across 15 game families"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T16:34:01.178138Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark of 140 Godot engine tasks across 15 game families. Coding agents must build playable games end-to-end, and the games are scored through replayed demonstrations and rubric-guided multimodal judging.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "evaluation protocol evidence", "public artifact URL"], "snippet": "We instantiate this framework as GameCraft-Bench, a benchmark comprising 140 Godot tasks across 15 game families."}, "familyId": "bmf_e53f3c332a7f", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_gamecraft-bench_5127c2aa", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2606.17861", "pdf": "https://arxiv.org/pdf/2606.17861", "project": "https://tongxuluo.github.io/gamecraft-bench-website", "report": "https://arxiv.org/abs/2606.17861"}, "metrics": [], "motivation": "Game generation is an emerging application of coding agents, requiring models to transform natural-language specifications into playable interactive systems.", "name": "GameCraft-Bench", "oneLine": "A benchmark of 140 Godot engine tasks across 15 game families. Coding agents must build playable games end-to-end, and the games are scored through replayed demonstrations and rubric-guided multimodal judging.", "paperTitle": "GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.17861", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.75, "relation": "introduces", "releasedAt": "2026-06-16", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Tongxu Luo", "Rongsheng Wang", "Jiaxi Bi", "Chenming Xu", "Zhengyang Tang", "Jianlong Chen", "Juhao Liang", "Ke Ji", "Shuqi Guo", "Yuhao Du", "Fan Bu", "Wenyu Du", "Xiaotong Zhang", "Kyle Li", "Shaobo Wang", "Linfeng Zhang", "Yuxuan Liu", "Xin Lai", "Chenxin Li", "Yiduo Guo", "Zhexin Zhang", "Xinyuan Wang", "Tianyi Bai", "Ziniu Li", "Benyou Wang"], "categories": ["cs.CL"], "id": "2606.17861", "title": "GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?", "type": "arxiv", "url": "https://arxiv.org/abs/2606.17861"}, "sourceUpdatedAt": "2026-06-16T12:34:39Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "It tests whether coding agents can produce complete, playable games inside a real engine, not isolated code snippets. The best reported agent scores 41.46%."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
