{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.26029", "kind": "paper", "url": "https://arxiv.org/abs/2606.26029"}], "authors": ["Yu-Yang Chen", "Lan-Zhe Guo"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.26029", "source_url": "https://arxiv.org/abs/2606.26029"}, "categories": ["General AI", "Geometric reasoning", "Reasoning", "Vision & 3D"], "description": {"en": "TriViewBench evaluates multimodal LLMs on multi-view structural reasoning using synthetic 3D scenes with controlled object count and occlusion. It comprises 1,923 scenes and over 14,000 QA pairs across four complexity levels and three reasoning categories: Local Decision, Object Counting, and Global Recovery."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.26029", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.26029", "title": "TriViewBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Nanjing University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.26029", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.26029", "source_url": "https://openalex.org/W7165962563"}, "key": "claire-radar:2606.26029", "modality": null, "name": "TriViewBench", "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_triviewbench_b0684034", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "965db6264bdb8727bfa01a3a0f684c9aa8937ef519e58ceae0642c99de4f3034", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.26029"}, "publisher": null, "released": "2026-06-24", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.26029", "source_url": "https://arxiv.org/abs/2606.26029"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-26029", "source": "claire_radar", "source_benchmark_id": "2606.26029", "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": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.26029", "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", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": ["Reasoning", "Geometric reasoning"], "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": "c359dc6ba00b933cad70c9512f2674719f6afa83b31e3a1712fbd3a0882a4e02", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "TriViewBench evaluates multimodal LLMs on multi-view structural reasoning using synthetic 3D scenes with controlled object count and occlusion. It comprises 1,923 scenes and over 14,000 QA pairs across four complexity levels and three reasoning categories: Local Decision, Object Counting, and Global Recovery.", "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"], "snippet": "We introduce TriViewBench, a controlled three-view visual reasoning benchmark constructed from synthetic 3D scenes with explicitly parameterized object count and occlusion."}, "familyId": "bmf_75312dbd377d", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_triviewbench_b0684034", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.26029", "paper": "https://arxiv.org/abs/2606.26029", "pdf": "https://arxiv.org/pdf/2606.26029", "project": null, "report": "https://arxiv.org/abs/2606.26029"}, "metrics": [], "motivation": "Multimodal Large Language Models (MLLMs) demonstrate strong performance on standard visual question answering benchmarks, yet their scalability under controlled structural complexity remains poorly understood.", "name": "TriViewBench", "oneLine": "TriViewBench evaluates multimodal LLMs on multi-view structural reasoning using synthetic 3D scenes with controlled object count and occlusion. It comprises 1,923 scenes and over 14,000 QA pairs across four complexity levels and three reasoning categories: Local Decision, Object Counting, and Global Recovery.", "paperTitle": "TriViewBench: Controlled Complexity Scaling for Multi-View Structural Reasoning in MLLMs", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.26029", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-24", "source": {"authors": ["Yu-Yang Chen and Lan-Zhe Guo"], "categories": ["cs.CV", "cs.AI"], "id": "2606.26029", "title": "TriViewBench: Controlled Complexity Scaling for Multi-View Structural Reasoning in MLLMs", "type": "arxiv", "url": "https://arxiv.org/abs/2606.26029"}, "sourceUpdatedAt": "2026-06-24T17:00:05Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "TriViewBench provides controlled complexity scaling to isolate structural reasoning capabilities in MLLMs, revealing distinct failure modes and bottlenecks. It enables systematic comparison of models on multi-view spatial reasoning, informing targeted improvements."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
