{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.17433", "kind": "paper", "url": "https://arxiv.org/abs/2606.17433"}, {"id": "hf:sahasrak/ladbench", "kind": "dataset", "url": "https://huggingface.co/datasets/SahasraK/LADBench"}], "authors": ["Sahasra Kondapalli", "Lara Radovanovic", "Aadi Palnitkar", "Mingyang Mao", "Xiaomin Lin"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.17433", "source_url": "https://arxiv.org/abs/2606.17433"}, "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "LADBench evaluates large vision-language models on detecting logical anomalies in synthetic images across four domains: Residential, Urban, Collaborative, and Nature. It uses a Tiered Prompting Protocol with three progressive disclosure levels and automated scoring."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.17433", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.17433", "title": "LADBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of South Florida"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.17433", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.17433", "source_url": "https://openalex.org/W7165023948"}, "key": "claire-radar:2606.17433", "modality": null, "name": "LADBench", "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_ladbench_44ed9c65", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "9e0a031dc831ed5bb810aa568b9d4329875145af08f3e9b515cd1b20e6b9a288", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.17433"}, "publisher": null, "released": "2026-06-16", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.17433", "source_url": "https://arxiv.org/abs/2606.17433"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-17433", "source": "claire_radar", "source_benchmark_id": "2606.17433", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Machine Generated", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": "SahasraK/LADBench", "hfDatasetDownloads": 309, "hfDatasetLikes": 2, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.17433", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-10-02", "state": "stale"}, "hfDatasetLikes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-10-02", "state": "stale"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "stale"}, "capabilities": [], "confidence": "High", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "f25d55dd9bb8f8885b5281bc5e8cf2438f1c259ac5418c1a484844da40b2d5fe", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "LADBench evaluates large vision-language models on detecting logical anomalies in synthetic images across four domains: Residential, Urban, Collaborative, and Nature. It uses a Tiered Prompting Protocol with three progressive disclosure levels and automated scoring.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "To address this, we introduce LAD-bench, a benchmark of more than 1,000 curated synthetic images with logical anomalies across four domains: Residential, Urban, Collaborative, and Nature."}, "familyId": "bmf_3e35e374df54", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_ladbench_44ed9c65", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": "https://huggingface.co/datasets/SahasraK/LADBench", "hfPaper": "https://huggingface.co/papers/2606.17433", "paper": "https://arxiv.org/abs/2606.17433", "pdf": "https://arxiv.org/pdf/2606.17433", "project": null, "report": "https://arxiv.org/abs/2606.17433"}, "metrics": [], "motivation": "Large Vision Language Models (VLMs) excel at visual question answering and semantic grounding, but their capacity for autonomous logical reasoning remains underexplored.", "name": "LADBench", "oneLine": "LADBench evaluates large vision-language models on detecting logical anomalies in synthetic images across four domains: Residential, Urban, Collaborative, and Nature. It uses a Tiered Prompting Protocol with three progressive disclosure levels and automated scoring.", "paperTitle": "LADBench: A Benchmark for Logical Fault Detection in Images", "primaryDomain": "General AI", "publication": {"evidence": "Accepted to the IEEE International Conference on Development and Learning (ICDL 2026)", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2606.17433", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "IEEE International Conference on Development and Learning (ICDL 2026)", "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "LADBench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://huggingface.co/datasets/SahasraK/LADBench"}], "ranking": {}, "readiness": "Inspectable", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-06-16", "source": {"authors": ["Sahasra Kondapalli", "Lara Radovanovic", "Aadi Palnitkar", "Mingyang Mao", "Xiaomin Lin"], "categories": ["cs.CV"], "id": "2606.17433", "title": "LADBench: A Benchmark for Logical Fault Detection in Images", "type": "arxiv", "url": "https://arxiv.org/abs/2606.17433"}, "sourceUpdatedAt": "2026-06-16T02:32:38Z", "topics": ["cs.CV"], "venueAttempts": [{"decisionRaw": "Accepted to the IEEE International Conference on Development and Learning (ICDL 2026)", "evidence": [{"level": "author-claim", "observedAt": "2026-10-14T01:46:58.058199Z", "rawValue": "Accepted to the IEEE International Conference on Development and Learning (ICDL 2026)", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2606.17433"}], "reviewStatus": "accepted", "venueName": "IEEE International Conference on Development and Learning (ICDL 2026)"}], "whyItMatters": "Existing anomaly benchmarks focus on visual errors, not the physical and social common sense required for open-world deployment. LADBench quantifies how much explicit assistance models need to localize and reason about logical faults, addressing a gap in evaluating sequential multimodal reasoning."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
