{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.15867", "kind": "paper", "url": "https://arxiv.org/abs/2606.15867"}], "authors": ["Long-Bao Nguyen", "Quang-Khai Le", "Tam V. Nguyen", "Minh-Triet Tran", "Trung-Nghia Le"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.15867", "source_url": "https://arxiv.org/abs/2606.15867"}, "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "CogCanvas is a benchmark for multi-subject reference-based image generation, with 1,952 curated reference images, 1,361 compositional prompts, and a unified six-axis evaluation protocol. It includes tasks for reference-based generation, text-to-image composition, and reference retrieval, with metrics BG-Sim and Attr-VQA."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.15867", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.15867", "title": "CogCanvas"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Vietnam National University Ho Chi Minh City", "Ho Chi Minh City University of Science", "University of Dayton"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.15867", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.15867", "source_url": "https://openalex.org/W7164897819"}, "key": "claire-radar:2606.15867", "modality": null, "name": "CogCanvas", "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_cogcanvas_d3de924a", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "66a99e3edb7c48de96a900a38c8c32748c8380c8052a444aa31c5f54408e0133", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.15867"}, "publisher": null, "released": "2026-06-14", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.15867", "source_url": "https://arxiv.org/abs/2606.15867"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-15867", "source": "claire_radar", "source_benchmark_id": "2606.15867", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "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/2606.15867", "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": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "399d38ad19fb4377b9470179b792d6a403db3f0c94795b3db9317e1669b49650", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "CogCanvas is a benchmark for multi-subject reference-based image generation, with 1,952 curated reference images, 1,361 compositional prompts, and a unified six-axis evaluation protocol. It includes tasks for reference-based generation, text-to-image composition, and reference retrieval, with metrics BG-Sim and Attr-VQA.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce CogCanvas, a benchmark of 1,952 curated reference images spanning 100 celebrity identities, 115 distinctive objects and fashion items, and 29 real-world background scenes including landmarks, from which we construct 1,361 compositional prompts covering 2-5 person group sizes."}, "familyId": "bmf_5fd28976da11", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_cogcanvas_d3de924a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.15867", "paper": "https://arxiv.org/abs/2606.15867", "pdf": "https://arxiv.org/pdf/2606.15867", "project": null, "report": "https://arxiv.org/abs/2606.15867"}, "metrics": [], "motivation": "Multi-subject reference-based image generation requires jointly preserving multiple human identities, binding per-person objects and fashion items, and respecting a specified background scene, a regime where current diffusion models remain brittle.", "name": "CogCanvas", "oneLine": "CogCanvas is a benchmark for multi-subject reference-based image generation, with 1,952 curated reference images, 1,361 compositional prompts, and a unified six-axis evaluation protocol. It includes tasks for reference-based generation, text-to-image composition, and reference retrieval, with metrics BG-Sim and Attr-VQA.", "paperTitle": "CogCanvas: A Benchmark for Evaluating Multi-Subject Reference-Based Image Generation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.15867", "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": "2026-06-14", "source": {"authors": ["Long-Bao Nguyen and Quang-Khai Tran and Tam V. Nguyen and Minh-Triet Tran and Trung-Nghia Le"], "categories": ["cs.CV"], "id": "2606.15867", "title": "CogCanvas: A Benchmark for Evaluating Multi-Subject Reference-Based Image Generation", "type": "arxiv", "url": "https://arxiv.org/abs/2606.15867"}, "sourceUpdatedAt": "2026-06-14T15:38:24Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "CogCanvas addresses the gap in jointly evaluating multi-identity, object binding, background grounding, and spatial plausibility in image generation. It provides a comprehensive benchmark for assessing composition capabilities across group sizes."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
