{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.03614", "kind": "paper", "url": "https://arxiv.org/abs/2607.03614"}, {"id": "gh:ideal3d/ideal-bench", "kind": "repo", "url": "https://github.com/IDEAL3D/IDEAL-Bench"}, {"kind": "website", "url": "https://ideal3d.github.io/"}], "authors": ["Yuening Cai", "Junwei Zhou", "Youran Qu", "Yu-Wing Tai"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.03614", "source_url": "https://arxiv.org/abs/2607.03614"}, "categories": ["General AI", "Geometric reasoning", "Reasoning", "Vision & 3D"], "description": {"en": "IDEAL-Bench evaluates Vision-Language Models on holistic 3D layout inference from single images of indoor scenes, scoring predictions across five numerical dimensions and a perceptual render-and-compare protocol. It uses a procedurally generated dataset of 1,000 re-renderable Blender environments."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.03614", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.03614", "title": "IDEAL-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.03614", "modality": null, "name": "IDEAL-Bench", "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_ideal-bench_dab673c0", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "a010415042b0aaf7e3e2f2ecac747cf2894e3a7d661c428abcaac827864cf8e3", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.03614"}, "publisher": null, "released": "2026-07-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.03614", "source_url": "https://arxiv.org/abs/2607.03614"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-03614", "source": "claire_radar", "source_benchmark_id": "2607.03614", "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": "https://github.com/IDEAL3D/IDEAL-Bench", "githubScope": "benchmark_repo", "githubStars": 4, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.03614", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "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": "3c2efb906e1db2c52ce9292af6a1fcfee7c4a6e02a7b8289ced2d1b156bb41f0", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "IDEAL-Bench evaluates Vision-Language Models on holistic 3D layout inference from single images of indoor scenes, scoring predictions across five numerical dimensions and a perceptual render-and-compare protocol. It uses a procedurally generated dataset of 1,000 re-renderable Blender environments.", "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", "public artifact URL"], "snippet": "To this end, we introduce IDEAL-Bench, an evaluation suite that requires VLMs to predict structured 3D layouts on photorealistic indoor scenes across 10 room types, scored along five numerical dimensions and a perceptual render-and-compare protocol."}, "familyId": "bmf_b9aa0b64dd23", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 1, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "linked-from-project", "linkedUrl": "https://github.com/IDEAL3D/IDEAL-Bench", "url": "https://ideal3d.github.io/"}], "scope": "benchmark_repo", "url": "https://github.com/IDEAL3D/IDEAL-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_ideal-bench_dab673c0", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/IDEAL3D/IDEAL-Bench", "data": null, "hfPaper": "https://huggingface.co/papers/2607.03614", "paper": "https://arxiv.org/abs/2607.03614", "pdf": "https://arxiv.org/pdf/2607.03614", "project": "https://ideal3d.github.io/", "report": "https://arxiv.org/abs/2607.03614"}, "metrics": [], "motivation": "Spatial question answering is the dominant paradigm for evaluating spatial intelligence in Vision-Language Models (VLMs), but it leaves a complementary axis of spatial competence under-evaluated: holistic 3D layout inference, which predicts every visible object's pose and extent from a single image in a structured form.", "name": "IDEAL-Bench", "oneLine": "IDEAL-Bench evaluates Vision-Language Models on holistic 3D layout inference from single images of indoor scenes, scoring predictions across five numerical dimensions and a perceptual render-and-compare protocol. It uses a procedurally generated dataset of 1,000 re-renderable Blender environments.", "paperTitle": "IDEAL-Bench: Indoor Dataset and Evaluation suite for Analyzing 3D Layout reasoning", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.03614", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "IDEAL-Bench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://ideal3d.github.io/"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-03", "source": {"authors": ["Yuening Cai", "Junwei Zhou", "Youran Qu", "Yu-Wing Tai"], "categories": ["cs.CV"], "id": "2607.03614", "title": "IDEAL-Bench: Indoor Dataset and Evaluation suite for Analyzing 3D Layout reasoning", "type": "arxiv", "url": "https://arxiv.org/abs/2607.03614"}, "sourceUpdatedAt": "2026-07-03T22:04:42Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Current VLM spatial evaluation relies on question answering, which misses structural understanding. IDEAL-Bench provides a reproducible, quantitative assessment of geometric and structural competencies, revealing model weaknesses in measuring scenes rather than describing them, and offering a diagnostic for genuine spatial intelligence."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
