{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.31529", "kind": "paper", "url": "https://arxiv.org/abs/2605.31529"}, {"id": "gh:texaser/svi-bench", "kind": "repo", "url": "https://github.com/Texaser/SVI-Bench"}], "authors": ["Yulu Pan", "Han Yi", "Seongsu Ha", "Md Mohaiminul Islam", "Benjamin Zhang", "Lorenzo Torresani", "Gedas Bertasius"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.31529", "source_url": "https://arxiv.org/abs/2605.31529"}, "categories": ["Agents", "General AI", "Vision & 3D"], "description": {"en": "SVI-Bench evaluates vision-language models on strategic video intelligence using sports as a microworld. It includes 9 tasks across 4 pillars: perception, reasoning, simulation, and agency, using basketball, soccer, and hockey videos with annotated actions and reports."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.31529", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.31529", "title": "SVI-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.31529", "modality": null, "name": "SVI-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": "unknown", "origin_record_id": "bm_svi-bench_918464db", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "87f8aa6dde69e2c27ba1b98b873f06c169a0fb923489452c8790ebbc5891f222", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.31529"}, "publisher": null, "released": "2026-05-29", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.31529", "source_url": "https://arxiv.org/abs/2605.31529"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-31529", "source": "claire_radar", "source_benchmark_id": "2605.31529", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Machine Generated", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/Texaser/SVI-Bench", "githubScope": "benchmark_repo", "githubStars": 2, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 5, "hfPaperUrl": "https://huggingface.co/papers/2605.31529", "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": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "97e6cb3853ee8765a4d34720db6c57d4d38cdd6b9cd140e15205201be13bfa7b", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "SVI-Bench evaluates vision-language models on strategic video intelligence using sports as a microworld. It includes 9 tasks across 4 pillars: perception, reasoning, simulation, and agency, using basketball, soccer, and hockey videos with annotated actions and reports.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To bridge this gap, we introduce SVI-Bench, a large-scale benchmark that leverages team sports as a dynamic microworld, combining the complexity of real-world multi-agent interaction (10-22 agents making coordinated decisions under adversarial pressure) with the verifiability of explicit rules and definitive outcomes."}, "familyId": "bmf_17a1b917580e", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/Texaser/SVI-Bench", "url": "https://github.com/Texaser/SVI-Bench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/Texaser/SVI-Bench", "url": "https://huggingface.co/papers/2605.31529"}], "scope": "benchmark_repo", "url": "https://github.com/Texaser/SVI-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_svi-bench_918464db", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/Texaser/SVI-Bench", "data": null, "hfPaper": "https://huggingface.co/papers/2605.31529", "paper": "https://arxiv.org/abs/2605.31529", "pdf": "https://arxiv.org/pdf/2605.31529", "project": null, "report": "https://arxiv.org/abs/2605.31529"}, "metrics": [], "motivation": "True video intelligence demands more than recognizing what is visible: it requires reasoning about why events unfold, predicting what would change under different conditions, and deciding what to do next.", "name": "SVI-Bench", "oneLine": "SVI-Bench evaluates vision-language models on strategic video intelligence using sports as a microworld. It includes 9 tasks across 4 pillars: perception, reasoning, simulation, and agency, using basketball, soccer, and hockey videos with annotated actions and reports.", "paperTitle": "SVI-Bench: A Dynamic Microworld for Strategic Video Intelligence", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.31529", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "MVP Group", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/Texaser/SVI-Bench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-29", "source": {"authors": ["Yulu Pan", "Han Yi", "Seongsu Ha", "Md Mohaiminul Islam", "Benjamin Zhang", "Lorenzo Torresani", "Gedas Bertasius"], "categories": ["cs.CV"], "id": "2605.31529", "title": "SVI-Bench: A Dynamic Microworld for Strategic Video Intelligence", "type": "arxiv", "url": "https://arxiv.org/abs/2605.31529"}, "sourceUpdatedAt": "2026-07-01T03:25:50Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Existing video benchmarks lack verifiable ground truth for causal and strategic reasoning. SVI-Bench combines real-world multi-agent complexity with verifiable rules and outcomes, enabling evaluation of higher-level cognitive capabilities in video understanding."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
