{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.08630", "kind": "paper", "url": "https://arxiv.org/abs/2608.08630"}, {"id": "gh:sharelab-sii/vlzip", "kind": "repo", "url": "https://github.com/ShareLab-SII/VLZip"}], "categories": ["General AI", "Long Context", "Reasoning", "Vision & 3D"], "description": {"en": "LongVLBench evaluates long-context multimodal understanding over interleaved visual and textual sequences, using tasks derived from video narratives that require holistic, narrative-level reasoning."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.08630", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.08630", "title": "LongVLBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.08630", "modality": null, "name": "LongVLBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "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_longvlbench_2560ef96", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "9fa9034c8a828c9fa9a53054d2d27fd370835bd96615af1badd6e15a2234fd97", "relation": "evaluates_only", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-03T09:03:47.731241Z", "source_url": "https://arxiv.org/abs/2608.08630"}, "publisher": null, "released": "2026-08-09", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.08630", "source_url": "https://arxiv.org/abs/2608.08630"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-08630", "source": "claire_radar", "source_benchmark_id": "2608.08630", "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": "https://github.com/ShareLab-SII/VLZip", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 6, "hfPaperUrl": "https://huggingface.co/papers/2608.08630", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The combination of a long-context multimodal benchmark and a new compression method addresses a current VLM scaling pain point, and the release includes accessible code and data.", "score": 72}, "capabilities": ["Reasoning"], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-03T09:03:47.731241Z", "inputHash": "c4f04022043ea472a2c0cd0d7925f167de5733b826a45f51e9c35a1461ea072b", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We also introduce LongVLBench, a new benchmark derived from video narratives that demands holistic, narrative-level reasoning.", "canonicalNameSource": "abstract", "decisionReason": "The benchmark is explicitly introduced with a public Hugging Face dataset, evaluation scripts, and a scoring setup, enabling other teams to run and score submissions.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-03T09:03:47.731241Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "LongVLBench evaluates long-context multimodal understanding over interleaved visual and textual sequences, using tasks derived from video narratives that require holistic, narrative-level reasoning.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "To address deficient evaluations in the field, we also introduce LongVLBench, a new benchmark derived from video narratives that demands holistic, narrative-level reasoning."}, "familyId": "bmf_c465aedb35ba", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_longvlbench_2560ef96", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/ShareLab-SII/VLZip", "data": null, "paper": "https://arxiv.org/abs/2608.08630", "pdf": "https://arxiv.org/pdf/2608.08630", "project": null, "report": "https://arxiv.org/abs/2608.08630"}, "metrics": [], "motivation": "Vision Language Models (VLMs) face significant challenges with ultra-long, interleaved image-text sequences due to the quadratic complexity of self-attention.", "name": "LongVLBench", "oneLine": "LongVLBench evaluates long-context multimodal understanding over interleaved visual and textual sequences, using tasks derived from video narratives that require holistic, narrative-level reasoning.", "paperTitle": "VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.08630", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "ShareLab, Fudan University and Shanghai Innovation Institute", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/ShareLab-SII/VLZip"}], "ranking": {"90d": {"confidence": "Medium", "coverage": 0.7, "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": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.34383183387990046, "value": 6}}, "confidence": "Medium", "coverage": 0.7, "rank": 549, "score": 20}, "method": "attention-ranking-v13", "rank": 549, "score": 20}}, "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-08-09", "source": {"authors": ["Yuqi Zhang", "Cheng Chen", "Yuyu Guo", "Wenjie Yang", "Lingchen Meng", "Peng Di", "Hang Yu", "Zuxuan Wu", "and Yu-Gang Jiang"], "categories": ["cs.CV"], "id": "2608.08630", "title": "VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling", "type": "arxiv", "url": "https://arxiv.org/abs/2608.08630"}, "sourceUpdatedAt": "2026-08-09T10:31:03Z", "topics": ["Long Context", "Reasoning"], "venueAttempts": [], "whyItMatters": "It addresses the lack of benchmarks for ultra-long interleaved image-text inputs, providing a public testbed for narrative reasoning beyond existing short-context multimodal evaluations."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
