{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.30256", "kind": "paper", "url": "https://arxiv.org/abs/2605.30256"}, {"kind": "website", "url": "https://research.nvidia.com/labs/amri/projects/video-fdb/"}], "authors": ["Amrita Mazumdar", "Seonwook Park", "Rajarshi Roy", "Nikhil Srihari", "Shengze Wang", "Yuhao Zhou", "Julia Wang", "Koki Nagano", "Shalini De Mello"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.30256", "source_url": "https://arxiv.org/abs/2605.30256"}, "categories": ["Cybersecurity", "Multimodal", "cs.CV"], "description": {"en": "VideoFDB is a benchmark presented for evaluating full-duplex audio-visual conversational agents. It includes 237 dyadic clips with 11 nonverbal conversational dynamics from real-world video calls, along with a taxonomy and rubric-based LM-as-judge evaluation framework."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.30256", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.30256", "title": "VideoFDB"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.30256", "modality": null, "name": "VideoFDB", "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_videofdb_d27be39c", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "5963744e9288c87bd2baf7e27e43ed802fde00da78d0d99539f56ea481274ea1", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.30256"}, "publisher": null, "released": "2026-05-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.30256", "source_url": "https://arxiv.org/abs/2605.30256"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-30256", "source": "claire_radar", "source_benchmark_id": "2605.30256", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Cybersecurity"], "area": "Multimodal", "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": 3, "hfPaperUrl": "https://huggingface.co/papers/2605.30256", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "status": "stale"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "0d5639bb80e973a363a731f3c7ef71dec4880a7a85703f699fcfe88472a4ddf0", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "VideoFDB is a benchmark presented for evaluating full-duplex audio-visual conversational agents. It includes 237 dyadic clips with 11 nonverbal conversational dynamics from real-world video calls, along with a taxonomy and rubric-based LM-as-judge evaluation framework.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "In this work, we present VideoFDB, the first benchmark to evaluate full-duplex audio-visual-to-audio-visual (AV2AV) conversational agents."}, "familyId": "bmf_e0d2574209dd", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_videofdb_d27be39c", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Cybersecurity"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2605.30256", "paper": "https://arxiv.org/abs/2605.30256", "pdf": "https://arxiv.org/pdf/2605.30256", "project": "https://research.nvidia.com/labs/amri/projects/video-fdb/", "report": "https://arxiv.org/abs/2605.30256"}, "metrics": [], "motivation": "Natural human conversation is full-duplex and audio-visual: people simultaneously speak and listen while continuously interpreting and producing nonverbal cues, such as nods, smiles, and gestures.", "name": "VideoFDB", "oneLine": "VideoFDB is a benchmark presented for evaluating full-duplex audio-visual conversational agents. It includes 237 dyadic clips with 11 nonverbal conversational dynamics from real-world video calls, along with a taxonomy and rubric-based LM-as-judge evaluation framework.", "paperTitle": "VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents", "primaryDomain": "Cybersecurity", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.30256", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Inspectable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-28", "source": {"authors": ["Amrita Mazumdar", "Seonwook Park", "Rajarshi Roy", "Nikhil Srihari", "Shengze Wang", "Yuhao Zhou", "Julia Wang", "Koki Nagano", "Shalini De Mello"], "categories": ["cs.CV", "cs.CL", "cs.HC"], "id": "2605.30256", "title": "VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents", "type": "arxiv", "url": "https://arxiv.org/abs/2605.30256"}, "sourceUpdatedAt": "2026-07-28T20:35:17Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "Existing full-duplex benchmarks only evaluate speech, missing the audio-visual nature of natural conversation. VideoFDB aims to fill this gap by evaluating agents that must produce and interpret nonverbal cues."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
