{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.03673", "kind": "paper", "url": "https://arxiv.org/abs/2609.03673"}, {"id": "gh:amap-ml/stateagent", "kind": "repo", "url": "https://github.com/AMAP-ML/StateAgent"}], "authors": ["Yingmao Miao", "Pengfei Zhang", "Chaoran Xu", "Meng Yu", "Jing Tang", "Xiangxiang Chu", "Chao Shen", "Chenhao Lin"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.03673", "source_url": "https://arxiv.org/abs/2609.03673"}, "categories": ["General AI", "Reasoning", "Vision & 3D"], "description": {"en": "Evaluates world-state reasoning in video continuation through 200 cross-segment tasks across past-visible, occluded-process, and complex-transition state categories, with scoring based on state equivalence, state correctness, and hallucination rate."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.03673", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.03673", "title": "Statebench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.03673", "modality": null, "name": "Statebench", "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_statebench_8e2244ba", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "057baa9825280593610ac199433b099014fa4a0306189d4360c8045260983070", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-07T00:42:15.394147Z", "source_url": "https://arxiv.org/abs/2609.03673"}, "publisher": null, "released": "2026-09-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.03673", "source_url": "https://arxiv.org/abs/2609.03673"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-03673", "source": "claire_radar", "source_benchmark_id": "2609.03673", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-09-04", "attemptedAt": "2026-09-04T06:36:46.783164Z", "githubRepo": "https://github.com/AMAP-ML/StateAgent", "githubScope": "benchmark_repo", "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.03673", "observedAt": "2026-09-03T06:34:20.331209Z", "signalStatus": {"githubStars": {"checkedAt": "2026-09-04T06:36:46.783164Z", "state": "unavailable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-04T06:36:46.783164Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Video generation evaluation is an active area and the benchmark offers a focused probe with clear artifacts, likely drawing interest from the video model and evaluation research communities.", "score": 58}, "capabilities": ["Reasoning"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-07T00:42:15.394147Z", "inputHash": "fd4ac07526d770c457ebe960420bf51742d6ceea4dca0fef70c22639d9cbe795", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "To answer this question, we introduce Statebench, a benchmark that targets this gap by testing continuations over three state categories", "canonicalNameSource": "abstract", "decisionReason": "The benchmark includes a downloadable dataset, evaluation scripts, and defined metrics, with public code and a Hugging Face dataset link, meeting reuse and scoring requirements.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-07T00:42:15.394147Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates world-state reasoning in video continuation through 200 cross-segment tasks across past-visible, occluded-process, and complex-transition state categories, with scoring based on state equivalence, state correctness, and hallucination rate.", "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", "public artifact URL"], "snippet": "To answer this question, we introduce Statebench, a benchmark that targets this gap by testing continuations over three state categories: past-visible states, occluded-process states, and complex-transition states."}, "familyId": "bmf_73bbcdc58298", "firstSeenAt": "2026-09-04", "heat": null, "id": "bm_statebench_8e2244ba", "indexedAt": "2026-09-04T06:13:54.443310Z", "industrySectors": [], "links": {"code": "https://github.com/AMAP-ML/StateAgent", "data": null, "paper": "https://arxiv.org/abs/2609.03673", "pdf": "https://arxiv.org/pdf/2609.03673", "project": null, "report": "https://arxiv.org/abs/2609.03673"}, "metrics": [], "motivation": "Video generators build long videos by composing shorter parts, either by generating segments one after another or by autoregressively extending chunks.", "name": "Statebench", "oneLine": "Evaluates world-state reasoning in video continuation through 200 cross-segment tasks across past-visible, occluded-process, and complex-transition state categories, with scoring based on state equivalence, state correctness, and hallucination rate.", "paperTitle": "Do Video Generators Track the World Across Segments? A Benchmark and Method for World-State Reasoning in Video Continuation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.03673", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-04T06:13:54.443310Z"}, "publications": [], "publishers": [{"name": "AMAP-ML", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/AMAP-ML/StateAgent"}], "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}, "today": {"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": 1, "historyDate": "2026-09-03", "rank": null, "score": null, "windowDays": 1}, "level": {"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}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.63, "role": "bonus", "value": 63}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": 16}, "method": "attention-ranking-v13", "rank": null, "score": 16}}, "readiness": "Runnable", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-09-03", "source": {"authors": ["Yingmao Miao", "Pengfei Zhang", "Chaoran Xu", "Meng Yu", "Jing Tang", "Xiangxiang Chu", "Chao Shen", "Chenhao Lin"], "categories": ["cs.CV"], "id": "2609.03673", "title": "Do Video Generators Track the World Across Segments? A Benchmark and Method for World-State Reasoning in Video Continuation", "type": "arxiv", "url": "https://arxiv.org/abs/2609.03673"}, "sourceUpdatedAt": "2026-09-03T11:09:46Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Provides a targeted test of whether video generators maintain consistent world state across segments, addressing a weakness in long-form video composition beyond frame-level quality."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
