{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.24652", "kind": "paper", "url": "https://arxiv.org/abs/2605.24652"}], "authors": ["Jialiang Yang", "Bin Xia", "Ruihang Chu", "Dingdong Wang", "Wanke Xia", "Zhun Mou", "Tianyang Zhong", "Yiting Zhao", "Wenming Yang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.24652", "source_url": "https://arxiv.org/abs/2605.24652"}, "categories": ["General AI", "Vision & 3D", "cs.AI"], "description": {"en": "AVBench evaluates audio-video generative models across ten fine-grained dimensions covering visual quality, audio quality, and cross-modal consistency for human-centric scenarios. It uses specialized evaluators trained via preference learning and provides continuous scores from prediction confidence."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.24652", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.24652", "title": "AVBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Tsinghua University", "Chinese University of Hong Kong", "Chinese University of Hong Kong, Shenzhen"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2605.24652", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2605.24652", "source_url": "https://openalex.org/W7162418187"}, "key": "claire-radar:2605.24652", "modality": null, "name": "AVBench", "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": "unknown", "origin_record_id": "bm_avbench_ce8c8830", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "a82b7078b2734184de57d87b0dc5665f470628a8b4c997ce42606941b6852d21", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.24652"}, "publisher": null, "released": "2026-05-23", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.24652", "source_url": "https://arxiv.org/abs/2605.24652"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-24652", "source": "claire_radar", "source_benchmark_id": "2605.24652", "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": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 1, "hfPaperUrl": "https://huggingface.co/papers/2605.24652", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": [], "confidence": "Medium", "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": "c2a081dc2cfddf81dcb28c0d41671c9feedc190ff6f47141701fe425db4215e7", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "AVBench evaluates audio-video generative models across ten fine-grained dimensions covering visual quality, audio quality, and cross-modal consistency for human-centric scenarios. It uses specialized evaluators trained via preference learning and provides continuous scores from prediction confidence.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence"], "snippet": "To address these issues, we introduce AVBench, a fully automated benchmark tailored for human-centric AV generation."}, "familyId": "bmf_edaffec7807d", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_avbench_ce8c8830", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2605.24652", "paper": "https://arxiv.org/abs/2605.24652", "pdf": "https://arxiv.org/pdf/2605.24652", "project": null, "report": "https://arxiv.org/abs/2605.24652"}, "metrics": [], "motivation": "Rapid advances in audio-video (AV) generation have enabled high-fidelity synthesis with synchronized sound, particularly for human-related scenarios involving speech and interactions.", "name": "AVBench", "oneLine": "AVBench evaluates audio-video generative models across ten fine-grained dimensions covering visual quality, audio quality, and cross-modal consistency for human-centric scenarios. It uses specialized evaluators trained via preference learning and provides continuous scores from prediction confidence.", "paperTitle": "AVBench: Human-Aligned and Automated Evaluation Benchmark for Audio-Video Generative Models", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.24652", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-05-23", "source": {"authors": ["Jialiang Yang and Bin Xia and Ruihang Chu and Dingdong Wang and Wanke Xia and Zhun Mou and Tianyang Zhong and Yiting Zhao and Wenming Yang"], "categories": ["cs.AI", "cs.CV", "cs.MM", "cs.SD"], "id": "2605.24652", "title": "AVBench: Human-Aligned and Automated Evaluation Benchmark for Audio-Video Generative Models", "type": "arxiv", "url": "https://arxiv.org/abs/2605.24652"}, "sourceUpdatedAt": "2026-05-23T16:42:39Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Existing benchmarks for AV generation are coarse and rely on generic multimodal LLMs, leading to inaccurate assessments. AVBench offers automated, human-aligned evaluation with fine-grained metrics, enabling reliable model comparison and serving as a potential reward signal for RLHF."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
