{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.01813", "kind": "paper", "url": "https://arxiv.org/abs/2607.01813"}, {"id": "gh:pris-cv/mmbench-live", "kind": "repo", "url": "https://github.com/PRIS-CV/MMBench-Live"}], "authors": ["Yuanzhi Liu", "Shousheng Zhao", "Bo Zhou", "Kongming Liang", "Zhanyu Ma"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.01813", "source_url": "https://arxiv.org/abs/2607.01813"}, "categories": ["Agents", "General AI", "Multimodal"], "description": {"en": "MMBench-Live is a continuously evolving multimodal benchmark built by a multi-agent pipeline from MMBench. It contains 5.9K newly generated evaluation instances with executable reasoning, and evaluates vision-language models across question-answer generation and reasoning tasks. Updates cost about USD 30 and take 1-2 hours, with distribution-consistent strategy to maintain comparability."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.01813", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.01813", "title": "MMBench-Live"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.01813", "modality": null, "name": "MMBench-Live", "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_mmbench-live_b26d79ee", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "01024610919d9558aac2bb71dbc6b4c88e4c007b6094438159d2a8288638cfce", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.01813"}, "publisher": null, "released": "2026-07-02", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.01813", "source_url": "https://arxiv.org/abs/2607.01813"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-01813", "source": "claire_radar", "source_benchmark_id": "2607.01813", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/PRIS-CV/MMBench-Live", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.01813", "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", "state": "unavailable"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "e87f2bd6c4e4ba1d586557018806e145d3d6044c1d41880947541bb3481736c4", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MMBench-Live is a continuously evolving multimodal benchmark built by a multi-agent pipeline from MMBench. It contains 5.9K newly generated evaluation instances with executable reasoning, and evaluates vision-language models across question-answer generation and reasoning tasks. Updates cost about USD 30 and take 1-2 hours, with distribution-consistent strategy to maintain comparability.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We present MMBench-Live, a continuously evolving multimodal benchmark built by a multi-agent-driven automated pipeline."}, "familyId": "bmf_23818de44e7f", "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/PRIS-CV/MMBench-Live", "url": "https://github.com/PRIS-CV/MMBench-Live"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/PRIS-CV/MMBench-Live", "url": "https://huggingface.co/papers/2607.01813"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/PRIS-CV/MMBench-Live", "url": "https://arxiv.org/abs/2607.01813"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/PRIS-CV/MMBench-Live.", "url": "https://arxiv.org/abs/2607.01813"}], "scope": "benchmark_repo", "url": "https://github.com/PRIS-CV/MMBench-Live", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_mmbench-live_b26d79ee", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/PRIS-CV/MMBench-Live", "data": null, "hfPaper": "https://huggingface.co/papers/2607.01813", "paper": "https://arxiv.org/abs/2607.01813", "pdf": "https://arxiv.org/pdf/2607.01813", "project": null, "report": "https://arxiv.org/abs/2607.01813"}, "metrics": [], "motivation": "Evaluation benchmarks are essential for assessing vision-language models (VLMs), but most multimodal benchmarks are static, making them vulnerable to temporal staleness, data contamination, and costly maintenance.", "name": "MMBench-Live", "oneLine": "MMBench-Live is a continuously evolving multimodal benchmark built by a multi-agent pipeline from MMBench. It contains 5.9K newly generated evaluation instances with executable reasoning, and evaluates vision-language models across question-answer generation and reasoning tasks. Updates cost about USD 30 and take 1-2 hours, with distribution-consistent strategy to maintain comparability.", "paperTitle": "MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.01813", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-02", "source": {"authors": ["Yuanzhi Liu", "Shousheng Zhao", "Bo Zhou", "Kongming Liang and Zhanyu Ma"], "categories": ["cs.CV", "cs.AI"], "id": "2607.01813", "title": "MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models", "type": "arxiv", "url": "https://arxiv.org/abs/2607.01813"}, "sourceUpdatedAt": "2026-07-02T07:27:58Z", "topics": ["Agents", "Multimodal"], "venueAttempts": [], "whyItMatters": "Static benchmarks suffer from contamination and staleness; this benchmark provides a scalable, low-cost paradigm for sustainable evaluation that preserves model rankings and reduces memorization signals."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
