{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.14189", "kind": "paper", "url": "https://arxiv.org/abs/2607.14189"}, {"id": "gh:zxhhh0201/multiref-compass", "kind": "repo", "url": "https://github.com/zxhhh0201/MultiRef-Compass"}], "authors": ["Xiaohan Zhang", "Yuqing Wen", "Junlin Chen", "Yuqi Tang", "Yiting He", "Lizhuo Shao", "Weiming Zhu", "Tengfei Liu", "Yang Shi", "Jialu Chen", "Yuanxing Zhang", "Huaxiong Li"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.14189", "source_url": "https://arxiv.org/abs/2607.14189"}, "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "MultiRef-Compass evaluates multi-reference-to-audio-video generation systems on 350 curated samples. It assesses Basic Quality, Reference Consistency, Audio-Visual Consistency, and Instruction Following using 14 sub-metrics, combining automatic metrics with an MLLM-as-a-Judge framework."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.14189", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.14189", "title": "MultiRef-Compass"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Nanjing University", "National University of Singapore", "The Hong Kong University of Science and Technology (Guangzhou)", "Tsinghua University", "Peking University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.14189", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.14189", "source_url": "https://openalex.org/W7169584414"}, "key": "claire-radar:2607.14189", "modality": null, "name": "MultiRef-Compass", "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_multiref-compass_396a4597", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "37a5584a9a8ef704e86961cc2562652bfac18ae385181ce0fe2530d9ff368a16", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.14189"}, "publisher": null, "released": "2026-07-15", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.14189", "source_url": "https://arxiv.org/abs/2607.14189"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-14189", "source": "claire_radar", "source_benchmark_id": "2607.14189", "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/zxhhh0201/MultiRef-Compass", "githubScope": "benchmark_repo", "githubStars": 13, "hfDailySubmittedAt": "2026-07-17T00:00:00.000Z", "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 30, "hfPaperUrl": "https://huggingface.co/papers/2607.14189", "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"}, "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": "56752da5440a76ba771efb3eed9a915e47f283e2c763da850ea795095b8f28a4", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MultiRef-Compass evaluates multi-reference-to-audio-video generation systems on 350 curated samples. It assesses Basic Quality, Reference Consistency, Audio-Visual Consistency, and Instruction Following using 14 sub-metrics, combining automatic metrics with an MLLM-as-a-Judge framework.", "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"], "snippet": "To address this gap, we introduce MultiRef-Compass, a unified benchmark for MR2AV generation."}, "familyId": "bmf_83d3a7c119e7", "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/zxhhh0201/MultiRef-Compass", "url": "https://github.com/zxhhh0201/MultiRef-Compass"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/zxhhh0201/MultiRef-Compass", "url": "https://huggingface.co/papers/2607.14189"}], "scope": "benchmark_repo", "url": "https://github.com/zxhhh0201/MultiRef-Compass", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_multiref-compass_396a4597", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/zxhhh0201/MultiRef-Compass", "data": null, "hfPaper": "https://huggingface.co/papers/2607.14189", "paper": "https://arxiv.org/abs/2607.14189", "pdf": "https://arxiv.org/pdf/2607.14189", "project": null, "report": "https://arxiv.org/abs/2607.14189"}, "metrics": [], "motivation": "Multi-reference-to-audio-video (MR2AV) generation aims to generate coherent audio-video content conditioned on multiple references and textual instructions.", "name": "MultiRef-Compass", "oneLine": "MultiRef-Compass evaluates multi-reference-to-audio-video generation systems on 350 curated samples. It assesses Basic Quality, Reference Consistency, Audio-Visual Consistency, and Instruction Following using 14 sub-metrics, combining automatic metrics with an MLLM-as-a-Judge framework.", "paperTitle": "MultiRef-Compass: Towards Comprehensive Evaluation of Multi-Reference-to-Audio-Video Generation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.14189", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "MultiRef-Compass Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/zxhhh0201/MultiRef-Compass"}], "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": 0.3522667134894853, "value": 13}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.606767028072414, "value": 30}}, "confidence": "Medium", "coverage": 0.7, "rank": 118, "score": 43}, "method": "attention-ranking-v13", "rank": 118, "score": 43}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-15", "source": {"authors": ["Xiaohan Zhang", "Yuqing Wen", "Junlin Chen", "Yuqi Tang", "Yiting He", "Lizhuo Shao", "Weiming Zhu", "Tengfei Liu", "Yang Shi", "Jialu Chen", "Yuanxing Zhang", "Huaxiong Li"], "categories": ["cs.CV", "cs.SD"], "id": "2607.14189", "title": "MultiRef-Compass: Towards Comprehensive Evaluation of Multi-Reference-to-Audio-Video Generation", "type": "arxiv", "url": "https://arxiv.org/abs/2607.14189"}, "sourceUpdatedAt": "2026-07-15T16:02:45Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "Existing benchmarks focus on text-driven or single-reference generation and often ignore joint audio-video alignment. MultiRef-Compass addresses the gap by providing a public protocol for multi-reference composition, enabling reproducible comparison across models in a rapidly evolving generation task."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
