{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.22516", "kind": "paper", "url": "https://arxiv.org/abs/2608.22516v1"}, {"kind": "website", "url": "https://buaa-colalab.github.io/TRACE/"}], "authors": ["Pengyiang Liu", "Junbo Niu", "Xiaoyang Hu", "Zhongyue Shi", "Zitian Wang", "Linjiang Huang", "Si Liu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.22516", "source_url": "https://arxiv.org/abs/2608.22516"}, "categories": ["General AI", "Information retrieval", "Vision & 3D", "cs.CV"], "description": {"en": "Tests long-horizon video understanding with 600 Temporal Ordering and Event Counting questions, auditing whether decoded frames cover all evidence intervals."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.22516v1", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.22516v1", "title": "VES-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.22516", "modality": null, "name": "VES-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": "true", "origin_record_id": "bm_ves-bench_b925ca24", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "7fb3e72cb63b2f35d2b9042248d9d1221414bc14cc6f9c4b99fb7dc4b875ca16", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-08-27T04:18:56.318436Z", "source_url": "https://arxiv.org/abs/2608.22516v1"}, "publisher": null, "released": "2026-08-23", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.22516", "source_url": "https://arxiv.org/abs/2608.22516v1"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-22516", "source": "claire_radar", "source_benchmark_id": "2608.22516", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-08-26", "attemptedAt": "2026-08-26T06:23:37.320575Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.22516", "observedAt": "2026-08-24T11:31:30.653116Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-08-26T06:23:37.320575Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "Novel audit-based evaluation for long video understanding may interest the video-language community, but missing artifact links in the input reduce immediate spread.", "score": 50}, "capabilities": ["Information retrieval"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:18:56.318436Z", "inputHash": "26230a94cb80423a590925fd33acf8f29148372b1dedaa39936d977aeee3866c", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce VES-Bench, a 600-question benchmark of Temporal Ordering and Event Counting items", "canonicalNameSource": "abstract", "decisionReason": "Formally named benchmark with a clear evaluation protocol and a project page, though no dataset or code links are supplied in the input.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:18:56.318436Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Tests long-horizon video understanding with 600 Temporal Ordering and Event Counting questions, auditing whether decoded frames cover all evidence intervals.", "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": "We introduce VES-Bench, a 600-question benchmark of Temporal Ordering and Event Counting items over 348 public long videos."}, "familyId": "bmf_23230c43b2ce", "firstSeenAt": "2026-08-25", "heat": null, "id": "bm_ves-bench_b925ca24", "indexedAt": "2026-08-25T17:37:25.889905Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "http://arxiv.org/abs/2608.22516v1", "pdf": "https://arxiv.org/pdf/2608.22516v1", "project": "https://buaa-colalab.github.io/TRACE/", "report": "http://arxiv.org/abs/2608.22516v1"}, "metrics": [], "motivation": "A long-video answer is evidence-supported only when the frames decoded from the video cover every event the answer depends on.", "name": "VES-Bench", "oneLine": "Tests long-horizon video understanding with 600 Temporal Ordering and Event Counting questions, auditing whether decoded frames cover all evidence intervals.", "paperTitle": "TRACE: Temporal Retrieval with Anchored and Convergent Evidence for Long-Horizon Video Understanding", "primaryDomain": "General AI", "publication": {"evidence": "Accepted to EMNLP 2026 Main Conference. 19 pages, 5 figures, 6 tables. Project page: https://buaa-colalab.github.io/TRACE/", "evidenceLevel": "author-claim", "evidenceUrl": "http://arxiv.org/abs/2608.22516v1", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP 2026 Main Conference", "verifiedAt": "2026-08-25T17:37:25.889905Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v12", "rank": null, "score": null}, "90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v12", "rank": null, "score": null}}, "readiness": "Paper only", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-08-23", "source": {"authors": ["Pengyiang Liu", "Junbo Niu", "Xiaoyang Hu", "Zhongyue Shi", "Zitian Wang", "Linjiang Huang", "Si Liu"], "categories": ["cs.CV", "cs.CL"], "id": "2608.22516", "title": "TRACE: Temporal Retrieval with Anchored and Convergent Evidence for Long-Horizon Video Understanding", "type": "arxiv", "url": "http://arxiv.org/abs/2608.22516v1"}, "sourceUpdatedAt": "2026-08-23T17:31:43Z", "topics": ["cs.CV"], "venueAttempts": [{"decisionRaw": "Accepted to EMNLP 2026 Main Conference. 19 pages, 5 figures, 6 tables. Project page: https://buaa-colalab.github.io/TRACE/", "evidence": [{"level": "author-claim", "observedAt": "2026-08-25T17:37:25.889905Z", "rawValue": "Accepted to EMNLP 2026 Main Conference. 19 pages, 5 figures, 6 tables. Project page: https://buaa-colalab.github.io/TRACE/", "sourceType": "arxiv-comments", "sourceUrl": "http://arxiv.org/abs/2608.22516v1"}], "reviewStatus": "accepted", "venueName": "EMNLP 2026 Main Conference"}], "whyItMatters": "Shifts evaluation from final answers to evidence coverage, revealing whether correct answers rest on complete observation of long videos."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
