{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.29074", "kind": "paper", "url": "https://arxiv.org/abs/2605.29074"}], "authors": ["Jiyao Zhang", "Mingxu Zhang", "Yitong Peng", "Haoxuan Liu", "Chenshuo Wang", "Yuxing Long", "Haoyang Huang", "Dongjiang Li", "Nan Duan", "Hui Shen", "Hao Dong"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.29074", "source_url": "https://arxiv.org/abs/2605.29074"}, "categories": ["Geometric reasoning", "Robotics", "Robotics & Embodied AI"], "description": {"en": "Embodied3DBench is a robot-centric benchmark for low-level spatial intelligence in embodied 3D environments. It includes 6 task categories: Grounding, Spatial Relation Prediction, Multi-view Correspondence, Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction, with 21k QA pairs across 12 subcategories."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.29074", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.29074", "title": "Embodied3DBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.29074", "modality": null, "name": "Embodied3DBench", "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_embodied3dbench_d183e439", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "27eab02b5dab49dad654072b290bf9cf2a680d04532ffee88cee8b23a9299896", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.29074"}, "publisher": null, "released": "2026-05-27", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.29074", "source_url": "https://arxiv.org/abs/2605.29074"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-29074", "source": "claire_radar", "source_benchmark_id": "2605.29074", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Robotics & Embodied AI"], "area": "Robotics & Embodied AI", "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/2605.29074", "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": ["Geometric reasoning"], "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": "49d23f33ffa8b41b6907b442b42771c8d6a65a7cab2d758b5f8fa1190032c601", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Embodied3DBench is a robot-centric benchmark for low-level spatial intelligence in embodied 3D environments. It includes 6 task categories: Grounding, Spatial Relation Prediction, Multi-view Correspondence, Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction, with 21k QA pairs across 12 subcategories.", "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 Embodied3DBench, a robot-centric benchmark targeting low-level spatial intelligence in embodied 3D environments."}, "familyId": "bmf_1a1023e37b98", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_embodied3dbench_d183e439", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Robotics"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2605.29074", "paper": "https://arxiv.org/abs/2605.29074", "pdf": "https://arxiv.org/pdf/2605.29074", "project": null, "report": "https://arxiv.org/abs/2605.29074"}, "metrics": [], "motivation": "Are current Vision Language Models (VLMs) ready to comprehend and reason about complex embodied interactions in 3D environments?", "name": "Embodied3DBench", "oneLine": "Embodied3DBench is a robot-centric benchmark for low-level spatial intelligence in embodied 3D environments. It includes 6 task categories: Grounding, Spatial Relation Prediction, Multi-view Correspondence, Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction, with 21k QA pairs across 12 subcategories.", "paperTitle": "Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models", "primaryDomain": "Robotics & Embodied AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.29074", "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": "2026-05-27", "source": {"authors": ["Jiyao Zhang", "Mingxu Zhang", "Yitong Peng", "Haoxuan Liu", "Chenshuo Wang", "Yuxing Long", "Haoyang Huang", "Dongjiang Li", "Nan Duan", "Hui Shen", "Hao Dong"], "categories": ["cs.CV", "cs.RO"], "id": "2605.29074", "title": "Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2605.29074"}, "sourceUpdatedAt": "2026-05-27T20:28:56Z", "topics": ["Robotics"], "venueAttempts": [], "whyItMatters": "Current VLMs show strong high-level spatial reasoning but lack interaction-oriented perception. Embodied3DBench reveals this gap and provides a scalable training dataset for improvement, enabling systematic evaluation and advancement of interaction-aware multimodal systems."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
