{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.28206", "kind": "paper", "url": "https://arxiv.org/abs/2608.28206v1"}], "authors": ["Sandeep Wadhwa", "Mayank Vatsa", "Richa Singh", "Parrva Chirag Shah", "Prakhar Galriya"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.28206", "source_url": "https://arxiv.org/abs/2608.28206"}, "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "Evaluates counting accuracy in text-to-image models using 640,000 prompts across 1,600 categories and counts from 1 to 100, with factorial variation of composition, layout, and appearance."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.28206v1", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.28206v1", "title": "NumBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Indian Institute of Technology Jodhpur", "Shiv Nadar University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.28206", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.28206", "source_url": "https://openalex.org/W7204760606"}, "key": "claire-radar:2608.28206", "modality": null, "name": "NumBench", "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": "true", "origin_record_id": "bm_numbench_cf80a48e", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "e050e9a75c37f5471fef593e0a35daf49e5874b957e1a1bda1128bd515a287e5", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-05T08:28:38.611381Z", "source_url": "https://arxiv.org/abs/2608.28206v1"}, "publisher": null, "released": "2026-08-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.28206", "source_url": "https://arxiv.org/abs/2608.28206v1"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-28206", "source": "claire_radar", "source_benchmark_id": "2608.28206", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Mixed", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-09-05", "attemptedAt": "2026-09-05T01:35:52.848019Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.28206", "observedAt": "2026-09-04T06:36:46.783164Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-05T01:35:52.848019Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "A large benchmark focused on a common text-to-image failure with a proposed calibrated metric is likely to attract interest from the vision and generative model community.", "score": 75}, "capabilities": [], "confidence": "High", "construction": "Aggregate Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-05T08:28:38.611381Z", "inputHash": "0df99bac3c2b96e4df3ed54cbcd4ca6024bf5716f2764a298f1d724d0e67343c", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce \\textbf{NumBench}, a benchmark of 640{,}000 prompts spanning 1{,}600 categories and counts from 1 to 100.", "canonicalNameSource": "abstract", "decisionReason": "The benchmark is formally named, defines a clear scoring metric, and the abstract indicates release of evaluation through proposal of a public score; the paper provides a public reuse path via arXiv and Hugging Face paper page.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-05T08:28:38.611381Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates counting accuracy in text-to-image models using 640,000 prompts across 1,600 categories and counts from 1 to 100, with factorial variation of composition, layout, and appearance.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce \\textbf{NumBench}, a benchmark of 640{,}000 prompts spanning 1{,}600 categories and counts from 1 to 100."}, "familyId": "bmf_f2a79fd9ccd2", "firstSeenAt": "2026-08-31", "heat": null, "id": "bm_numbench_cf80a48e", "indexedAt": "2026-08-31T07:22:53.604319Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2608.28206", "paper": "http://arxiv.org/abs/2608.28206v1", "pdf": "https://arxiv.org/pdf/2608.28206v1", "project": null, "report": "http://arxiv.org/abs/2608.28206v1"}, "metrics": [], "motivation": "Text-to-image (T2I) models often generate the wrong number of objects, yet existing benchmarks are too small or weakly controlled to explain why.", "name": "NumBench", "oneLine": "Evaluates counting accuracy in text-to-image models using 640,000 prompts across 1,600 categories and counts from 1 to 100, with factorial variation of composition, layout, and appearance.", "paperTitle": "NumBench: Diagnosing Counting Failures in Text-to-Image Models", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "http://arxiv.org/abs/2608.28206v1", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-05T06:02:50.019181Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}, "90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-28", "source": {"authors": ["Sandeep Wadhwa", "Mayank Vatsa", "Richa Singh", "Parrva Chirag Shah", "Prakhar Galriya"], "categories": ["cs.CV", "cs.DB"], "id": "2608.28206", "title": "NumBench: Diagnosing Counting Failures in Text-to-Image Models", "type": "arxiv", "url": "http://arxiv.org/abs/2608.28206v1"}, "sourceUpdatedAt": "2026-08-28T11:26:19Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "Provides a large-scale controlled setting to diagnose counting failures and introduces a calibrated scoring metric that discounts uncertain detections, enabling rigorous model comparison."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
