{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.05264", "kind": "paper", "url": "https://arxiv.org/abs/2607.05264"}], "authors": ["Suryanarayana Reddy Yarrabothula", "Manisha Chawla", "Kunal Sinha", "Gagan Raj Gupta", "Sashank Lekkala", "Ashirvadhan Dosapati", "Saikamal Nannuri", "Katragadda Ajay RamaSwamy Chowdary Gowtham"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.05264", "source_url": "https://arxiv.org/abs/2607.05264"}, "categories": ["General AI", "Multimodal", "Reasoning", "Safety", "Safety & Trustworthiness"], "description": {"en": "STEELBENCH evaluates vision-language models on per-worker activity recognition and safety-rule reasoning in industrial CCTV footage. It includes 1,345 clips with dense annotations and a provenance-aware audit protocol."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.05264", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.05264", "title": "SteelBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Indian Institute of Technology Bhilai", "Vellore Institute of Technology University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.05264", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.05264", "source_url": "https://openalex.org/W7167640782"}, "key": "claire-radar:2607.05264", "modality": null, "name": "SteelBench", "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_steelbench_6bf9f31b", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "f63ca45281ea9f379799d2e6bf635b3fde59bd52c23e1a7d9abd027b0a4d191f", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.05264"}, "publisher": null, "released": "2026-07-06", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.05264", "source_url": "https://arxiv.org/abs/2607.05264"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-05264", "source": "claire_radar", "source_benchmark_id": "2607.05264", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "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/2607.05264", "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": ["Reasoning"], "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": "b95fe3291dce8586dd5090da725478438b20b1e36996f31f3e4c6369a81b8646", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "STEELBENCH evaluates vision-language models on per-worker activity recognition and safety-rule reasoning in industrial CCTV footage. It includes 1,345 clips with dense annotations and a provenance-aware audit protocol.", "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": "We introduce STEELBENCH, a diagnostic benchmark for industrial surveillance that jointly evaluates per-worker activity recognition, safety-rule reasoning, and annotation provenance."}, "familyId": "bmf_c0bc763f6367", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_steelbench_6bf9f31b", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2607.05264", "paper": "https://arxiv.org/abs/2607.05264", "pdf": "https://arxiv.org/pdf/2607.05264", "project": null, "report": "https://arxiv.org/abs/2607.05264"}, "metrics": [], "motivation": "Existing video benchmarks evaluate action recognition on consumer videos, egocentric recordings, or simulated industrial environments.", "name": "SteelBench", "oneLine": "STEELBENCH evaluates vision-language models on per-worker activity recognition and safety-rule reasoning in industrial CCTV footage. It includes 1,345 clips with dense annotations and a provenance-aware audit protocol.", "paperTitle": "SteelBench: Evaluating Vision-Language Models in Real-World Industrial Environments", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.05264", "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": "2026-07-06", "source": {"authors": ["Suryanarayana Reddy Yarrabothula", "Manisha Chawla", "Kunal Sinha", "Gagan Raj Gupta", "Sashank Lekkala", "Ashirvadhan Dosapati", "Saikamal Nannuri", "Katragadda Ajay RamaSwamy Chowdary Gowtham"], "categories": ["cs.CV"], "id": "2607.05264", "title": "SteelBench: Evaluating Vision-Language Models in Real-World Industrial Environments", "type": "arxiv", "url": "https://arxiv.org/abs/2607.05264"}, "sourceUpdatedAt": "2026-07-06T16:11:45Z", "topics": ["Multimodal", "Safety", "Reasoning"], "venueAttempts": [], "whyItMatters": "Industrial surveillance poses unique visual and procedural challenges. This benchmark reveals large gaps in VLM performance and shows how annotation provenance can inflate accuracy, guiding reliable deployment."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
