{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.26641", "kind": "paper", "url": "https://arxiv.org/abs/2605.26641"}, {"kind": "website", "url": "https://yunzeliu.github.io/OmniRetriever/"}], "authors": ["Yunze Liu", "Chi-Hao Wu", "Enmin Zhou", "Junxiao Shen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.26641", "source_url": "https://arxiv.org/abs/2605.26641"}, "categories": ["General AI", "Information retrieval", "Vision & 3D", "cs.CV"], "description": {"en": "A released 12-direction audio-video-text retrieval benchmark containing 3,782 triples and reporting aggregate retrieval performance."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.26641", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.26641", "title": "OmniRetriever-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.26641", "modality": null, "name": "OmniRetriever-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_omniretriever-bench_9cb59982", "origin_source": "arxiv", "recognition_confidence": "0.6", "record_sha256": "c74c1a59bff9e5c779e6d850a47a0218f820e81c16ddea73dae7c0eb8894d11c", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T18:00:24.082641Z", "source_url": "https://arxiv.org/abs/2605.26641"}, "publisher": null, "released": "2026-05-26", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.26641", "source_url": "https://arxiv.org/abs/2605.26641"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-26641", "source": "claire_radar", "source_benchmark_id": "2605.26641", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The any-to-any multimodal retrieval setting and comparison with prominent embedding systems may generate near-term attention.", "score": 3}, "capabilities": ["Information retrieval"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T18:00:24.082641Z", "inputHash": "909030d67a4b8a0adfb053e7fec3c767074a74e7f3be6e3dd4d1fdb0ddf4782c", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "we further release OmniRetriever-Bench", "canonicalNameSource": "abstract", "decisionReason": "Publish because the abstract formally names and releases the benchmark, specifies its retrieval task and scale, and reports a comparable AVG-all score.", "evaluationEvidence": {"reuse": "we further release OmniRetriever-Bench", "scoring": "OmniRetriever-7B attains AVG-all 34.84", "task": "a 12-direction AVT retrieval benchmark totaling 3782 triples"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T18:00:24.082641Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A released 12-direction audio-video-text retrieval benchmark containing 3,782 triples and reporting aggregate retrieval performance.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "public artifact URL"], "snippet": "To stress-test joint representations, we further release OmniRetriever-Bench, a 12-direction AVT retrieval benchmark totaling 3782 triples; on it OmniRetriever-7B attains AVG-all 34.84, improving over Gemini Embedding 2 by 1.72 and over the best prior open-source AVT method by 8.03."}, "familyId": "bmf_5e7f80e3af14", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_omniretriever-bench_9cb59982", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2605.26641", "pdf": "https://arxiv.org/pdf/2605.26641", "project": "https://yunzeliu.github.io/OmniRetriever/", "report": "https://arxiv.org/abs/2605.26641"}, "metrics": [], "motivation": "Unified multimodal embedding spaces have become the standard interface for cross-modal retrieval and multimodal RAG, and recent audio-video-text (AVT) encoders extend this setting to three modalities.", "name": "OmniRetriever-Bench", "oneLine": "A released 12-direction audio-video-text retrieval benchmark containing 3,782 triples and reporting aggregate retrieval performance.", "paperTitle": "OmniRetriever: Any-to-Any Audio-Video-Text Retrieval via Fusion-as-Teacher Distillation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.26641", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.6, "relation": "introduces", "releasedAt": "2026-05-26", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Yunze Liu", "Chi-Hao Wu", "Enmin Zhou", "Junxiao Shen"], "categories": ["cs.CV"], "id": "2605.26641", "title": "OmniRetriever: Any-to-Any Audio-Video-Text Retrieval via Fusion-as-Teacher Distillation", "type": "arxiv", "url": "https://arxiv.org/abs/2605.26641"}, "sourceUpdatedAt": "2026-05-26T07:26:23Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "It evaluates joint representations across retrieval directions involving audio, video, and text rather than limiting evaluation to pairwise retrieval."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
