{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.01245", "kind": "paper", "url": "https://arxiv.org/abs/2607.01245"}], "authors": ["Firoz Shaik", "Mateus Picanço Lima Gomes", "Tanvir Aumi", "Jingci Wang", "Milos Milunovic", "Filip Basara", "Ivana Jovanovic", "Vishwas Suryanarayanan", "Neha Nandan Kenkare", "Weiyao Xie", "Zhipeng Han", "Zheng Zhang", "Waleed Shahid", "Jay Rathi", "Russell Scherer", "Thong Q. Nguyen", "Michael Bentley", "Tamara Stankovic", "Rasika Chakravarthy", "Vishal Chowdhary"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.01245", "source_url": "https://arxiv.org/abs/2607.01245"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "A two-track benchmark evaluating structural, visual, and domain-grounded comprehension of native Word, Excel, and PowerPoint files."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.01245", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.01245", "title": "Office Comprehension Benchmark"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.01245", "modality": null, "name": "Office Comprehension Benchmark", "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_office-comprehension-benchmark_81bf3f7a", "origin_source": "arxiv", "recognition_confidence": "0.6", "record_sha256": "10b4939f15f9b9d4fcc5c9c3425cfaa3971e3ea6061065a56c963213235c2fd0", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T18:09:08.870777Z", "source_url": "https://arxiv.org/abs/2607.01245"}, "publisher": null, "released": "2026-05-29", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.01245", "source_url": "https://arxiv.org/abs/2607.01245"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-01245", "source": "claire_radar", "source_benchmark_id": "2607.01245", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The supplied text establishes public benchmark infrastructure but provides no evidence suitable for a confident short-term attention forecast.", "score": 50}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T18:09:08.870777Z", "inputHash": "91655defb0b859d4fbe0dea0efc06cb3feeb4d6109b93fa9491adcf4443108f5", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "Office Comprehension Benchmark", "canonicalNameSource": "paper_title", "decisionReason": "The source formally names the benchmark, defines repeatable evaluation tasks and claim-level scoring, and states that the dataset, evaluation tooling, judge prompt, and leaderboard are public.", "evaluationEvidence": {"reuse": "We release the dataset, evaluation tooling, judge prompt, and a public leaderboard.", "scoring": "Each reference answer is decomposed into atomic, binary-gradable claims, and an ensemble of LLM judges scores responses against each claim independently.", "task": "jointly evaluate LLM systems on Word, Excel, and PowerPoint comprehension over native file formats"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T18:09:08.870777Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A two-track benchmark evaluating structural, visual, and domain-grounded comprehension of native Word, Excel, and PowerPoint files.", "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"], "snippet": "We introduce Office Comprehension Bench (OCB), the first public benchmark to jointly evaluate LLM systems on Word, Excel, and PowerPoint comprehension over native file formats (.docx, .xlsx, .pptx) and their variants."}, "familyId": "bmf_174be76880f7", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_office-comprehension-benchmark_81bf3f7a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2607.01245", "pdf": "https://arxiv.org/pdf/2607.01245", "project": null, "report": "https://arxiv.org/abs/2607.01245"}, "metrics": [], "motivation": "We introduce Office Comprehension Bench (OCB), the first public benchmark to jointly evaluate LLM systems on Word, Excel, and PowerPoint comprehension over native file formats (.docx, .xlsx, .pptx) and their variants.", "name": "Office Comprehension Benchmark", "oneLine": "A two-track benchmark evaluating structural, visual, and domain-grounded comprehension of native Word, Excel, and PowerPoint files.", "paperTitle": "Office Comprehension Benchmark", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.01245", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.6, "relation": "introduces", "releasedAt": "2026-05-29", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Firoz Shaik", "Mateus Pican\\c{c}o Lima Gomes", "Tanvir Aumi", "Jingci Wang", "Milos Milunovic", "Filip Basara", "Ivana Jovanovic", "Vishwas Suryanarayanan", "Neha Nandan Kenkare", "Weiyao Xie", "Zhipeng Han", "Zheng Zhang", "Waleed Shahid", "Jay Rathi", "Russell Scherer", "Thong Q. Nguyen", "Michael Bentley", "Tamara Stankovic", "Rasika Chakravarthy", "Vishal Chowdhary"], "categories": ["cs.CL", "cs.AI", "cs.CY", "cs.IR", "cs.LG"], "id": "2607.01245", "title": "Office Comprehension Benchmark", "type": "arxiv", "url": "https://arxiv.org/abs/2607.01245"}, "sourceUpdatedAt": "2026-05-29T03:19:32Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "It provides a reusable evaluation of office-document comprehension across native formats and professional domains, supported by claim-level grading tools and a public leaderboard."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
