{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.00218", "kind": "paper", "url": "https://arxiv.org/abs/2607.00218"}], "authors": ["Siddhant Panpatil", "Arth Singh", "Mijin Koo", "Chaeyun Kim", "Haon Park", "Dasol Choi"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.00218", "source_url": "https://arxiv.org/abs/2607.00218"}, "categories": ["Robotics", "Robotics & Embodied AI", "Safety"], "description": {"en": "EgoSafetyBench is a diagnostic egocentric video benchmark of 1,200 robot-view scenarios to evaluate vision-language models as runtime safety guards. It assesses situational awareness across routine, suspicious, obvious, and contextual hazards, and visual-channel robustness against misleading in-scene text."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.00218", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.00218", "title": "EgoSafetyBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Seoul National University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.00218", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.00218", "source_url": "https://openalex.org/W7167033496"}, "key": "claire-radar:2607.00218", "modality": null, "name": "EgoSafetyBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "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_egosafetybench_4d21ada3", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "f56d75ae1326e53831d23c0ba30ad372cc557401ec833184cf6f52d97869f83d", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.00218"}, "publisher": null, "released": "2026-06-30", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.00218", "source_url": "https://arxiv.org/abs/2607.00218"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-00218", "source": "claire_radar", "source_benchmark_id": "2607.00218", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Robotics & Embodied AI"], "area": "Robotics & Embodied AI", "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": 1, "hfPaperUrl": "https://huggingface.co/papers/2607.00218", "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", "lastSuccessfulDate": "2026-10-04", "state": "stale"}}, "status": "stale"}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "98c489eb2d4216f5b04646e759a36801531ad2874cb0d51bf52f0fc32cd8c5a9", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "EgoSafetyBench is a diagnostic egocentric video benchmark of 1,200 robot-view scenarios to evaluate vision-language models as runtime safety guards. It assesses situational awareness across routine, suspicious, obvious, and contextual hazards, and visual-channel robustness against misleading in-scene text.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence"], "snippet": "We introduce EgoSafetyBench, an egocentric video benchmark of 1,200 robot-view scenarios annotated at half-second granularity, to evaluate VLMs as streaming guards across two tracks."}, "familyId": "bmf_4c45392c5812", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_egosafetybench_4d21ada3", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Robotics"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2607.00218", "paper": "https://arxiv.org/abs/2607.00218", "pdf": "https://arxiv.org/pdf/2607.00218", "project": null, "report": "https://arxiv.org/abs/2607.00218"}, "metrics": [], "motivation": "Vision-language models (VLMs) are now proposed as runtime safety guards for embodied agents in homes and factories.", "name": "EgoSafetyBench", "oneLine": "EgoSafetyBench is a diagnostic egocentric video benchmark of 1,200 robot-view scenarios to evaluate vision-language models as runtime safety guards. It assesses situational awareness across routine, suspicious, obvious, and contextual hazards, and visual-channel robustness against misleading in-scene text.", "paperTitle": "EgoSafetyBench: A Diagnostic Egocentric Video Benchmark for Evaluating Embodied VLMs as Runtime Safety Guards", "primaryDomain": "Robotics & Embodied AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.00218", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-06-30", "source": {"authors": ["Siddhant Panpatil", "Arth Singh", "Mijin Koo", "Chaeyun Kim", "Haon Park", "Dasol Choi"], "categories": ["cs.CV", "cs.AI"], "id": "2607.00218", "title": "EgoSafetyBench: A Diagnostic Egocentric Video Benchmark for Evaluating Embodied VLMs as Runtime Safety Guards", "type": "arxiv", "url": "https://arxiv.org/abs/2607.00218"}, "sourceUpdatedAt": "2026-06-30T21:50:49Z", "topics": ["Robotics", "Safety"], "venueAttempts": [], "whyItMatters": "This benchmark addresses the practical need for safety guards that distinguish genuine hazards from superficially alarming but benign actions, and highlights the vulnerability to misleading signage, which can inform safer deployment of embodied VLMs in real-world settings."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
