{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.14660", "kind": "paper", "url": "https://arxiv.org/abs/2607.14660"}, {"id": "gh:mcg-nju/viabench", "kind": "repo", "url": "https://github.com/MCG-NJU/VIABench"}], "authors": ["Yunfeng Liu", "Yuandong Yang", "Jiarui Han", "Zhenpeng Huang", "Yuqing Tang", "Xiangyu Zeng", "Gangshan Wu", "Limin Wang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.14660", "source_url": "https://arxiv.org/abs/2607.14660"}, "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "VIABench evaluates multimodal large language models on three tasks from first-person videos of visually impaired individuals: proactive reminder, visual question answering, and vision-guided interaction. It includes 761 videos, 46.9 hours, and 14,526 annotations, with protocols for online and offline settings."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.14660", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.14660", "title": "VIABench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.14660", "modality": null, "name": "VIABench", "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_viabench_7ade7668", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "66563740915e1b84536e6e8cbfa1995d406016880c0e5a724e876a8b7bb3fd84", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.14660"}, "publisher": null, "released": "2026-07-16", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.14660", "source_url": "https://arxiv.org/abs/2607.14660"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-14660", "source": "claire_radar", "source_benchmark_id": "2607.14660", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/MCG-NJU/VIABench", "githubScope": "benchmark_repo", "githubStars": 8, "hfDailySubmittedAt": "2026-07-17T00:00:00.000Z", "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 6, "hfPaperUrl": "https://huggingface.co/papers/2607.14660", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "d1ad5b80f3052a709b1912c1c26133d7314ca31648473995fab20925fa140868", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "VIABench evaluates multimodal large language models on three tasks from first-person videos of visually impaired individuals: proactive reminder, visual question answering, and vision-guided interaction. It includes 761 videos, 46.9 hours, and 14,526 annotations, with protocols for online and offline settings.", "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", "public artifact URL"], "snippet": "To fill this gap, we introduce VIABench, a comprehensive video benchmark specifically designed to evaluate MLLMs in Visually Impaired Assistance scenarios using first-person videos recorded or shared by VIIs themselves."}, "familyId": "bmf_3bab85546b0e", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/MCG-NJU/VIABench", "url": "https://github.com/MCG-NJU/VIABench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/MCG-NJU/VIABench", "url": "https://huggingface.co/papers/2607.14660"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/MCG-NJU/VIABench.", "url": "https://arxiv.org/abs/2607.14660"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/MCG-NJU/VIABench", "url": "https://arxiv.org/abs/2607.14660"}], "scope": "benchmark_repo", "url": "https://github.com/MCG-NJU/VIABench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_viabench_7ade7668", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/MCG-NJU/VIABench", "data": null, "hfPaper": "https://huggingface.co/papers/2607.14660", "paper": "https://arxiv.org/abs/2607.14660", "pdf": "https://arxiv.org/pdf/2607.14660", "project": null, "report": "https://arxiv.org/abs/2607.14660"}, "metrics": [], "motivation": "Visually impaired individuals (VIIs) encounter significant daily challenges due to limited access to visual information.", "name": "VIABench", "oneLine": "VIABench evaluates multimodal large language models on three tasks from first-person videos of visually impaired individuals: proactive reminder, visual question answering, and vision-guided interaction. It includes 761 videos, 46.9 hours, and 14,526 annotations, with protocols for online and offline settings.", "paperTitle": "VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.14660", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "MCG-NJU", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/MCG-NJU/VIABench"}], "ranking": {"90d": {"confidence": "Medium", "coverage": 0.7, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.29328998349929525, "value": 8}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.34383183387990046, "value": 6}}, "confidence": "Medium", "coverage": 0.7, "rank": 169, "score": 36}, "method": "attention-ranking-v13", "rank": 169, "score": 36}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-16", "source": {"authors": ["Yunfeng Liu", "Yuandong Yang", "Jiarui Han", "Zhenpeng Huang", "Yuqing Tang", "Xiangyu Zeng", "Gangshan Wu", "Limin Wang"], "categories": ["cs.CV"], "id": "2607.14660", "title": "VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance", "type": "arxiv", "url": "https://arxiv.org/abs/2607.14660"}, "sourceUpdatedAt": "2026-07-16T07:25:17Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "General MLLMs are rarely tested for real-world assistance of blind users. VIABench focuses on tasks like anticipating navigation-critical events, which are underrepresented in existing benchmarks, providing a practical measure of model utility in assistive contexts."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
