{"record": {"aliases": ["CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage"], "artifacts": [{"id": "arxiv:2606.21618", "kind": "paper", "url": "https://arxiv.org/abs/2606.21618"}, {"id": "gh:zevtsao/culmind", "kind": "repo", "url": "https://github.com/ZevTsao/CulMind"}], "authors": ["Zhangwei Cao", "Shuhan Fan", "Yuting Wei", "Jiajun Zhang", "Yihang Peng", "Qi Meng", "Yangfu Zhu", "Liangbin Yang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.21618", "source_url": "https://arxiv.org/abs/2606.21618"}, "categories": ["General AI", "Multimodal", "Reasoning"], "description": {"en": "CulMind evaluates multimodal understanding across Chinese cultural-heritage tasks, while CulMind-R and ReaScore assess reasoning processes."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.21618", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.21618", "title": "CulMind"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of International Relations", "Kedge Business School", "Peking University", "Tsinghua University", "Capital Normal University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.21618", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.21618", "source_url": "https://openalex.org/W7165664089"}, "key": "claire-radar:2606.21618", "modality": null, "name": "CulMind", "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_culmind-benchmarking-multimodal-understand_7f4a50cb", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "58ca28f84339fc275d68facaeaef6fe433ab5fe646af72edf99541357250e574", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T17:16:57.488613Z", "source_url": "https://arxiv.org/abs/2606.21618"}, "publisher": null, "released": "2026-06-19", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.21618", "source_url": "https://arxiv.org/abs/2606.21618"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-21618", "source": "claire_radar", "source_benchmark_id": "2606.21618", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Its broad museum coverage and reasoning-focused evaluation may interest multimodal and cultural-computing researchers.", "score": 70}, "capabilities": ["Reasoning"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T17:16:57.488613Z", "inputHash": "c6f6998376354428262b75755c487215c4775a1bd8bab6e9a128f79875c4a6ac", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage", "canonicalNameSource": "paper_title", "decisionReason": "The source provides formal benchmark names, defined task coverage, a named scoring metric, and publicly released data, code, and evaluation scripts.", "evaluationEvidence": {"reuse": "We publicly release the data, code, and evaluation scripts at https://github.com/ZevTsao/CulMind", "scoring": "ReaScore, a task-adaptive metric that evaluates reasoning by automatically weighting task-relevant dimensions", "task": "a high-quality benchmark for multimodal CCH covering 50 tasks from collections of more than 100 museums"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T17:16:57.488613Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "CulMind evaluates multimodal understanding across Chinese cultural-heritage tasks, while CulMind-R and ReaScore assess reasoning processes.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "To address this gap, we introduce CulMind and CulMind-R: a high-quality benchmark for multimodal CCH covering 50 tasks from collections of more than 100 museums, and a 24-task reasoning subset that adaptively defines task-specific dimensions for reasoning process evaluation."}, "familyId": "bmf_187cd0c3c2e9", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_culmind-benchmarking-multimodal-understand_7f4a50cb", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/ZevTsao/CulMind", "data": null, "paper": "https://arxiv.org/abs/2606.21618", "pdf": "https://arxiv.org/pdf/2606.21618", "project": null, "report": "https://arxiv.org/abs/2606.21618"}, "metrics": [], "motivation": "Evaluating Multimodal Large Language Models (MLLMs) in Chinese Cultural Heritage (CCH) requires fine-grained reasoning over visual, textual, stylistic, and historical clues.", "name": "CulMind", "oneLine": "CulMind evaluates multimodal understanding across Chinese cultural-heritage tasks, while CulMind-R and ReaScore assess reasoning processes.", "paperTitle": "CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.21618", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-06-19", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Zhangwei Cao", "Shuhan Fan", "Yuting Wei", "Jiajun Zhang", "Yihang Peng", "Qi Meng", "Yangfu Zhu", "Liangbin Yang"], "categories": ["cs.CL"], "id": "2606.21618", "title": "CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage", "type": "arxiv", "url": "https://arxiv.org/abs/2606.21618"}, "sourceUpdatedAt": "2026-06-19T17:22:08Z", "topics": ["Multimodal", "Reasoning"], "venueAttempts": [], "whyItMatters": "It extends cultural-heritage evaluation beyond final-answer accuracy to task-specific reasoning quality."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
