{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.27187", "kind": "paper", "url": "https://arxiv.org/abs/2606.27187"}], "authors": ["Jiajun Wu", "Haoyu Kang", "Yining Sun", "Jiacheng Hou", "Heng Zhang", "Danyang Zhang", "Zhenjun Zhao", "Haochi Zhang", "Leixin Sun", "Eric Hanchen Jiang", "Yushan Li", "Ruiyu Li", "Mengkai Huang", "Yan Gao", "Xu Zhang", "Guancheng Wan"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.27187", "source_url": "https://arxiv.org/abs/2606.27187"}, "categories": ["General AI", "Multimodal"], "description": {"en": "Evaluates harmful video understanding in large multimodal models using 1,379 videos and 4,137 multiple-choice questions across three hierarchical dimensions: Observable Evidence, Clip-Internal Meaning, and Beyond-Clip Reasoning."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.27187", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.27187", "title": "HarmVideoBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.27187", "modality": null, "name": "HarmVideoBench", "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_harmvideobench_cc5648e1", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "2b860c53c3bbbf19a14465062335d7820f4e7755c4db37d76d297b58088c6124", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.27187"}, "publisher": null, "released": "2026-06-25", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.27187", "source_url": "https://arxiv.org/abs/2606.27187"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-27187", "source": "claire_radar", "source_benchmark_id": "2606.27187", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "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/2606.27187", "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": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "da19fcc6a36e1ba6b9e1ab2fdf885087bfa948216d6cc4da0be732640dff2ffe", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates harmful video understanding in large multimodal models using 1,379 videos and 4,137 multiple-choice questions across three hierarchical dimensions: Observable Evidence, Clip-Internal Meaning, and Beyond-Clip Reasoning.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To address these problems, we present HarmVideoBench, a multi-layered diagnostic benchmark comprising 1,379 videos paired with 4,137 multiple-choice questions."}, "familyId": "bmf_7d36f4d3eff5", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_harmvideobench_cc5648e1", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.27187", "paper": "https://arxiv.org/abs/2606.27187", "pdf": "https://arxiv.org/pdf/2606.27187", "project": null, "report": "https://arxiv.org/abs/2606.27187"}, "metrics": [], "motivation": "Large vision-language models (LVLMs) have recently shown immense potential in automated content moderation, sparking growing interest in developing harmful-video benchmarks.", "name": "HarmVideoBench", "oneLine": "Evaluates harmful video understanding in large multimodal models using 1,379 videos and 4,137 multiple-choice questions across three hierarchical dimensions: Observable Evidence, Clip-Internal Meaning, and Beyond-Clip Reasoning.", "paperTitle": "HarmVideoBench: Benchmarking Harmful Video Understanding in Large Multimodal Models", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.27187", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-25", "source": {"authors": ["Jiajun Wu", "Haoyu Kang", "Yining Sun", "Jiacheng Hou", "Heng Zhang", "Danyang Zhang", "Zhenjun Zhao", "Haochi Zhang", "Leixin Sun", "Eric Hanchen Jiang", "Yushan Li", "Ruiyu Li", "Mengkai Huang", "Yan Gao", "Xu Zhang", "Guancheng Wan"], "categories": ["cs.CV", "cs.CL"], "id": "2606.27187", "title": "HarmVideoBench: Benchmarking Harmful Video Understanding in Large Multimodal Models", "type": "arxiv", "url": "https://arxiv.org/abs/2606.27187"}, "sourceUpdatedAt": "2026-06-25T15:50:33Z", "topics": ["Multimodal"], "venueAttempts": [], "whyItMatters": "Addresses the gap of shallow binary classification in harmful video benchmarks by testing deep contextual understanding, and the absence of explanatory rationales, providing a diagnostic tool for model evaluation."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
