{"record": {"aliases": ["Auto-AEG: Scalable Data Construction for Open-Vocabulary Audio Event Grounding"], "artifacts": [{"id": "arxiv:2607.04383", "kind": "paper", "url": "https://arxiv.org/abs/2607.04383"}, {"id": "hf:zihan-audio/aegbench", "kind": "dataset", "url": "https://huggingface.co/datasets/zihan-audio/AEGBench"}], "authors": ["Zihan Zhang", "Xize Cheng", "Wenhao Yan", "Tong Zhang", "Dongjie Fu", "Wenxu Jia", "Boyun Zhang", "Yongbo He", "Qifei Zhang", "Linjun Li", "Wenshi Chen", "Tao Jin"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.04383", "source_url": "https://arxiv.org/abs/2607.04383"}, "categories": ["General AI", "Speech & Audio", "cs.SD"], "description": {"en": "AEGBench is a difficulty-stratified evaluation for open-vocabulary audio event grounding, where systems localize queried sound events in time."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.04383", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.04383", "title": "AEGBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Zhejiang University", "Tsinghua University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.04383", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.04383", "source_url": "https://openalex.org/W7167599887"}, "key": "claire-radar:2607.04383", "modality": null, "name": "AEGBench", "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_auto-aeg-scalable-data-construction-for-op_47893158", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "d8e62716b179b82d293a0aba8750b93cdb43cf52c1147ed75575dd9a89634162", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T17:26:02.308374Z", "source_url": "https://arxiv.org/abs/2607.04383"}, "publisher": null, "released": "2026-07-05", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.04383", "source_url": "https://arxiv.org/abs/2607.04383"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-04383", "source": "claire_radar", "source_benchmark_id": "2607.04383", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Auto-AEG: Scalable Data Construction for Open-Vocabulary Audio Event Grounding"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Speech & Audio", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The combination of audio-language models, open-vocabulary grounding, and a released dataset may attract specialist interest, although the work is marked as in progress.", "score": 69}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T17:26:02.308374Z", "inputHash": "c1dd09d4f5c0748b17b1321cda94ee197151e446bcd98b9c871c76eb7b114032", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "on AEGBench, an independent difficulty-stratified benchmark we release", "canonicalNameSource": "abstract", "decisionReason": "The supplied abstract formally names and releases AEGBench, defines its localization task, reports mIoU scoring, and supplies a public dataset path.", "evaluationEvidence": {"reuse": "AEGBench: https://huggingface.co/datasets/zihan-audio/AEGBench", "scoring": "+73.9% and +23.1% mIoU over zero-shot", "task": "predicting all time intervals of a target sound event described by an arbitrary natural language query"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T17:26:02.308374Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "AEGBench is a difficulty-stratified evaluation for open-vocabulary audio event grounding, where systems localize queried sound events in time.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "Training with this pipeline yields large temporal-localization gains (+73.9% and +23.1% mIoU over zero-shot) on AEGBench, an independent difficulty-stratified benchmark we release, and these gains generalize to held-out SED and other audio grounding benchmarks."}, "familyId": "bmf_253425773a8a", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_auto-aeg-scalable-data-construction-for-op_47893158", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": "https://huggingface.co/datasets/zihan-audio/AEGBench", "paper": "https://arxiv.org/abs/2607.04383", "pdf": "https://arxiv.org/pdf/2607.04383", "project": null, "report": "https://arxiv.org/abs/2607.04383"}, "metrics": [], "motivation": "Large Audio-Language Models (LALMs) reason fluently about sound yet struggle to localize precisely when events occur, while classical Sound Event Detection attains frame-level precision only over a closed label set.", "name": "AEGBench", "oneLine": "AEGBench is a difficulty-stratified evaluation for open-vocabulary audio event grounding, where systems localize queried sound events in time.", "paperTitle": "Auto-AEG: Scalable Data Construction for Open-Vocabulary Audio Event Grounding", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.04383", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "zihan-audio", "organizationType": "community", "role": "benchmark-publisher", "sourceUrl": "https://huggingface.co/datasets/zihan-audio/AEGBench"}], "readiness": "Inspectable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-07-05", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Zihan Zhang", "Xize Cheng", "Wenhao Yan", "Tong Zhang", "Dongjie Fu", "Boyun Zhang", "Yongbo He", "Tao Jin"], "categories": ["cs.SD", "cs.AI"], "id": "2607.04383", "title": "Auto-AEG: Scalable Data Construction for Open-Vocabulary Audio Event Grounding", "type": "arxiv", "url": "https://arxiv.org/abs/2607.04383"}, "sourceUpdatedAt": "2026-07-29T12:04:35Z", "topics": ["cs.SD"], "venueAttempts": [], "whyItMatters": "It provides reusable temporal-grounding data and an interval-overlap metric for comparing audio-language models beyond closed-label sound-event detection."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
