{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.11242", "kind": "paper", "url": "https://arxiv.org/abs/2609.11242"}, {"id": "hf:klingteam/vwg-bench", "kind": "dataset", "url": "https://huggingface.co/datasets/KlingTeam/VWG-Bench"}], "authors": ["Meng Luo", "Yicheng Liu", "Jiahao Wang", "Yuanxing Zhang", "Xin Tao", "Pengfei Wan", "Kun Gai", "Hao Fei"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.11242", "source_url": "https://arxiv.org/abs/2609.11242"}, "categories": ["General AI", "Reasoning", "Vision & 3D"], "description": {"en": "VWG-Bench evaluates image-to-video generation models on rule-following, physical and commonsense constraints, and goal realization over time, using 380 annotated initial images across 9 reasoning dimensions and 38 ten-sample task groups, scored with bundled 1–5 metrics and an evaluation toolkit."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.11242", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.11242", "title": "VWG-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.11242", "modality": null, "name": "VWG-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "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_vwg-bench_fa3de0ff", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "4c345dac8ce45a8ce0855e6be7c2080044bd33cb5133a5a14c1d396daee1e23d", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-14T00:43:53.625753Z", "source_url": "https://arxiv.org/abs/2609.11242"}, "publisher": null, "released": "2026-09-10", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.11242", "source_url": "https://arxiv.org/abs/2609.11242"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-11242", "source": "claire_radar", "source_benchmark_id": "2609.11242", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-09-11", "attemptedAt": "2026-09-11T12:23:23.267513Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": "KlingTeam/VWG-Bench", "hfDatasetDownloads": 406, "hfDatasetLikes": 1, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.11242", "observedAt": "2026-09-11T12:23:23.267513Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"observedAt": "2026-09-11T12:23:23.267513Z", "state": "fresh"}, "hfDatasetLikes": {"observedAt": "2026-09-11T12:23:23.267513Z", "state": "fresh"}, "hfPaperUpvotes": {"checkedAt": "2026-09-11T12:23:23.267513Z", "state": "unavailable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark addresses a topical video-generation reasoning area with a released Hugging Face dataset and runnable evaluation toolkit, though the 380-sample scale and niche image-to-video focus temper expected early attention.", "score": 54}, "capabilities": ["Reasoning"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-15T09:37:10.086794Z", "inputHash": "df2d32af35944d5acbe35865ff6b06650f02d7ae63a97fb360aae0ffa010c362", "model": "deepseek-v4-flash"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We introduce VWG-Bench (Video World Generalist Benchmark), a comprehensive benchmark", "canonicalNameSource": "abstract", "decisionReason": "The Hugging Face dataset includes annotations, evaluation code, and scoring metrics, providing a fixed public protocol for reuse.", "evaluationEvidence": {"reuse": "The test split contains 380 initial images and structured annotations ... The `evaluation_code/` directory contains the cleaned VWG-Bench evaluation toolkit and helper scripts.", "scoring": "The bundled `evaluation_code/` package evaluates five 1–5 metrics when applicable: 1. video quality; 2. progress consistency; 3. implicit-rule following; 4. progress-goal realization; 5. last-frame-goal realization.", "task": "VWG-Bench (Video World Generalist Benchmark) is an image-to-video reasoning benchmark for evaluating whether video generation models can follow symbolic rules, respect physical and commonsense constraints, and realize intended goals over time."}, "model": "deepseek-v4-pro", "reviewedAt": "2026-09-14T00:43:53.625753Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "VWG-Bench evaluates image-to-video generation models on rule-following, physical and commonsense constraints, and goal realization over time, using 380 annotated initial images across 9 reasoning dimensions and 38 ten-sample task groups, scored with bundled 1–5 metrics and an evaluation toolkit.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce VWG-Bench (Video World Generalist Benchmark), a comprehensive benchmark spanning 9 reasoning dimensions and 38 fine-grained tasks."}, "familyId": "bmf_6b6465e00623", "firstSeenAt": "2026-09-11", "heat": null, "id": "bm_vwg-bench_fa3de0ff", "indexedAt": "2026-09-11T00:48:15.671466Z", "industrySectors": [], "links": {"code": null, "data": "https://huggingface.co/datasets/KlingTeam/VWG-Bench", "paper": "https://arxiv.org/abs/2609.11242", "pdf": "https://arxiv.org/pdf/2609.11242", "project": null, "report": "https://arxiv.org/abs/2609.11242"}, "metrics": [], "motivation": "Video generation has advanced to produce visually compelling and temporally coherent results.", "name": "VWG-Bench", "oneLine": "VWG-Bench evaluates image-to-video generation models on rule-following, physical and commonsense constraints, and goal realization over time, using 380 annotated initial images across 9 reasoning dimensions and 38 ten-sample task groups, scored with bundled 1–5 metrics and an evaluation toolkit.", "paperTitle": "From Evaluation to Enhancement: Benchmarking and Improving Think-with-Video Reasoning for Video Generative Models", "primaryDomain": "General AI", "publication": {"evidence": "Accepted to ECCV 2026. 46 pages, 41 figures", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2609.11242", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "ECCV 2026", "verifiedAt": "2026-09-11T00:48:15.671466Z"}, "publications": [], "publishers": [{"name": "KlingTeam", "organizationType": "company-research-lab", "role": "benchmark-publisher", "sourceUrl": "https://huggingface.co/datasets/KlingTeam/VWG-Bench"}], "ranking": {"30d": {"confidence": "Low", "coverage": 0.15, "datasetDownloadRank": 11, "datasetRankPopulation": 42, "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": 0.6151846071448699, "value": 406}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.15, "rank": 22, "score": 52}, "method": "attention-ranking-v13", "rank": 22, "score": 52}, "90d": {"confidence": "Low", "coverage": 0.3, "datasetDownloadRank": 24, "datasetRankPopulation": 77, "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": 0.4981448575060773, "value": 406}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.3, "rank": 61, "score": 50}, "method": "attention-ranking-v13", "rank": 61, "score": 50}, "today": {"confidence": "Low", "coverage": 0.05, "datasetDownloadRank": 1, "datasetRankPopulation": 2, "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": 1, "historyDate": "2026-09-10", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": 0.75, "percentile": 0.75, "value": 406}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.55, "role": "bonus", "value": 55}}, "confidence": "Low", "coverage": 0.05, "rank": 3, "score": 53}, "method": "attention-ranking-v13", "rank": 3, "score": 53}}, "readiness": "Inspectable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-10", "source": {"authors": ["Meng Luo", "Yicheng Liu", "Jiahao Wang", "Yuanxing Zhang", "Xin Tao", "Pengfei Wan", "Kun Gai", "Hao Fei"], "categories": ["cs.CV", "cs.AI"], "id": "2609.11242", "title": "From Evaluation to Enhancement: Benchmarking and Improving Think-with-Video Reasoning for Video Generative Models", "type": "arxiv", "url": "https://arxiv.org/abs/2609.11242"}, "sourceUpdatedAt": "2026-09-10T08:40:58Z", "topics": ["Reasoning"], "venueAttempts": [{"decisionRaw": "Accepted to ECCV 2026. 46 pages, 41 figures", "evidence": [{"level": "author-claim", "observedAt": "2026-09-11T00:48:15.671466Z", "rawValue": "Accepted to ECCV 2026. 46 pages, 41 figures", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2609.11242"}], "reviewStatus": "accepted", "venueName": "ECCV 2026"}], "whyItMatters": "Video generation models are increasingly judged on whether they follow instructions and constraints over time, yet comparable image-to-video reasoning evaluation is limited; a released test split with a fixed metric suite and runnable evaluation code lets independent teams score their own generations on the same terms."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
