{"record": {"aliases": ["A Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation"], "artifacts": [{"id": "arxiv:2605.31351", "kind": "paper", "url": "https://arxiv.org/abs/2605.31351"}, {"id": "gh:yiyiyizhao/viable", "kind": "repo", "url": "https://github.com/YiyiyiZhao/VIABLE"}], "authors": ["Yi Zhao", "Siqi Wang", "Zhe Hu", "Yushi Li", "Jing Li"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.31351", "source_url": "https://arxiv.org/abs/2605.31351"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "A benchmark of over 300K judgment samples that evaluates VLM judges for visually impaired assistance on effectiveness, impartiality, and stability."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.31351", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.31351", "title": "VIABLE"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.31351", "modality": null, "name": "VIABLE", "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_a-visually-impaired-assistance-benchmark-f_ddc0146a", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "2c96f39b6cc2fe9eb4f462015becc4ae55bc5aefb191b2f13bf6ecb39498cb36", "relation": "introduces", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T16:50:07.444368Z", "source_url": "https://arxiv.org/abs/2605.31351"}, "publisher": null, "released": "2026-05-29", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.31351", "source_url": "https://arxiv.org/abs/2605.31351"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-31351", "source": "claire_radar", "source_benchmark_id": "2605.31351", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["A Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "Niche accessibility domain, but the LLM-as-judge reliability angle has broad interest.", "score": 40}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T16:50:07.444368Z", "inputHash": "5c5822aedf0c39cc4795f13f2bd6cf822573ffadf15b4bdda186e59ea0a561de", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "VIABLE (Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation)", "canonicalNameSource": "abstract", "decisionReason": "Named benchmark with a defined three-axis scoring framework and failure taxonomy, plus public data and code.", "evaluationEvidence": {"reuse": "Data and code are available at: https://github.com/YiyiyiZhao/VIABLE", "scoring": "introduces an Effectiveness--Impartiality--Stability framework with a 12-mode failure taxonomy", "task": "the first benchmark for VLM-as-a-Judge evaluation in VIA"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T16:50:07.444368Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark of over 300K judgment samples that evaluates VLM judges for visually impaired assistance on effectiveness, impartiality, and stability.", "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 investigate this question, we introduce VIABLE (Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation), the first benchmark for VLM-as-a-Judge evaluation in VIA."}, "familyId": "bmf_0a0dafe6a9fc", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_a-visually-impaired-assistance-benchmark-f_ddc0146a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/YiyiyiZhao/VIABLE", "data": null, "paper": "https://arxiv.org/abs/2605.31351", "pdf": "https://arxiv.org/pdf/2605.31351", "project": null, "report": "https://arxiv.org/abs/2605.31351"}, "metrics": [], "motivation": "AI-based Visually Impaired Assistance (VIA) remains challenging, largely due to the high cost of human evaluation.", "name": "VIABLE", "oneLine": "A benchmark of over 300K judgment samples that evaluates VLM judges for visually impaired assistance on effectiveness, impartiality, and stability.", "paperTitle": "A Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.31351", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-05-29", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Yi Zhao", "Siqi Wang", "Zhe Hu", "Yushi Li", "Jing Li"], "categories": ["cs.CL", "cs.CV"], "id": "2605.31351", "title": "A Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation", "type": "arxiv", "url": "https://arxiv.org/abs/2605.31351"}, "sourceUpdatedAt": "2026-05-29T14:28:20Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "It tests whether VLM-as-a-Judge can replace costly human evaluation in an accessibility domain, measuring diagnostic accuracy, self-preference, and adversarial fragility."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
